How is this ranking ordered?

Text models use the AIHOT v15 algorithm on ChinaAPI's curated model roster and verified source observations: fixed-weight pairwise comparisons followed by Kemeny consensus. Missing source weights are not redistributed. Image, video and scene views retain separate source metrics.

Why is a newly released model missing or near the bottom?

A model needs evidence from at least 3 sources, 3 families, 3 operators, 3 specialty categories and 2 direct anchors, within the release window. Models lacking evidence remain visible without a numerical rank. Unranked does not mean weak.

Does ChinaAPI run its own evaluation?

ChinaAPI performs no benchmark tests. It calculates text consensus ranks from cited public results. Original observations, configurations, dates and source links remain available.

How often is the snapshot updated?

Each update is a reviewed change that moves the snapshot date; the current snapshot is 2026-10-05 and the roster was last changed on 2026-10-05.

Which of these models can I call through ChinaAPI?

85 of the 103 models listed are available; each of those rows links to the model's catalogue page with its price.

Chinese AI Model Rankings

Name: ChinaAPI model capability rankings
Creator: ChinaAPI

As of 2026-10-05, Claude Opus 5.5 leads the ChinaAPI Consensus v15 at 81.8; the highest-placed Chinese model is Kimi K3 at 60.5 (#14). 85 of the 103 models listed can be called through ChinaAPI.

Where the Chinese AI models ChinaAPI serves stand next to the international frontier, quoted from public leaderboards. Snapshot of 2026-10-05; roster last maintained 2026-10-05 (moves when a model or source is added). Every figure is reproduced from the board named in its column and links back to it. ChinaAPI runs no evaluation of its own. Text uses the open-source AIHOT v15 consensus algorithm on ChinaAPI's curated roster; media and scene columns retain their source-specific metrics. See how this list is built, the price of every model marked available, and the same data as JSON at /api/model-rankings.

Sources

ChinaAPI Consensus v15

44 eligible models; available budget 62%. Missing weights are not redistributed. Admission requires 3 sources, 3 evaluation families, 3 operators, 3 specialty categories and 2 direct anchors within an 18-month release window. Insufficient evidence means unranked, not last. Scores are display values, not accuracy percentages. This uses a different candidate and evidence pool from AIHOT.

artificial-analysis: 30.00% (unavailable)
artificial-analysis-multilingual: 6.00% (reserved)
Arena Text Style-Controlled: 5.00% (available)
Arena WebDev: 2.40% (available)
Mercor APEX-Agents 1.1: 4.00% (available)
Vals Finance Agent: 3.00% (available)
DeepSWE v1.1: 3.60% (available)
TapTap Maker Benchmark: 3.60% (available)
LiveBench · 编程综合: 2.40% (available)
LiveBench · 推理与数学: 4.20% (available)
LiveBench · 语言与指令: 3.00% (available)
Arena Creative Writing: 5.00% (available)
Arena Vision Style-Controlled: 6.00% (available)
tau-banking: 2.00% (reserved)
SimpleQA Verified: 4.00% (available)
GPQA Diamond: 2.00% (available)
FrontierMath v2 · Tiers 1–3: 3.60% (available)
FrontierMath v2 · Tier 4: 1.20% (available)
Chess Puzzles: 1.80% (available)
Mystery Game Puzzles: 1.20% (available)
Creative Writing v3: 3.60% (available)
Longform Writing: 2.40% (available)

AIHOT v15 algorithm (MIT)

Text and multimodal models

Ordered by ChinaAPI Consensus v15. Unranked entries have no numerical position; their discovery order follows release date.

Text and multimodal models, public benchmark results as of 2026-10-05

Model

Vendor

Origin

Released

On ChinaAPI

ChinaAPI Consensus v15

Consensus evidence

LMArena Text Arena

LiveBench

Artificial Analysis Intelligence Index

SuperCLUE 智能指数

Epoch Capabilities Index (ECI)

Vals Index

Vendor's own evaluation card

LMArena Text Arena · Chinese

LMArena Text Arena · Coding

LiveBench · Coding

LiveBench · Agentic Coding

Artificial Analysis · Output Speed

Artificial Analysis · GDPval-AA v2.1

Artificial Analysis · Terminal-Bench v4.0

Artificial Analysis · MLCR-AA

Notes

New model watch — released in the 30 days ending 2026-10-05; awaiting evidence, no overall rank

—

Gemini 4 Argon

Google

International

2026-09-30

—

Unranked: insufficient families. 5 sources, 2 families, 3 operators; 19.4% budget coverage.

Source observations

Arena Text Style-Controlled: 1525.2150000155266 — gemini-4-argon-high; first_party:high@gemini-4-argon-high; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1679.5004220949027 — gemini-4-argon-high; first_party:high@gemini-4-argon-high; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 82.2 — gemini-4-argon; scaffolded:harness-loop-truncated-tools-agent+high+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gemini-4-argon; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 65.401 — google/gemini-4-argon; first_party:high@google-gemini-4-argon; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1518.826089395385 — gemini-4-argon-high; first_party:high@gemini-4-argon-high; retrieved 2026-10-04T14:23:50.120Z

Compare an earlier model: Gemini 3.8 Flash

Same evaluation protocol only; configurations are stated for each entry. No overall rank is inferred.

Source	Gemini 4 Argon	Gemini 3.8 Flash
Arena Text Style-Controlled	1525.22 — gemini-4-argon-high; High 推理; retrieved 2026-10-04T14:23:50.120Z	1494.77 — gemini-3.8-flash-high; High 推理; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev	1679.50 — gemini-4-argon-high; High 推理; retrieved 2026-10-04T14:23:50.120Z	1582.68 — gemini-3.8-flash-high; High 推理; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1	82.20 — gemini-4-argon; High 推理 · loop_truncated_tools_agent · apex-agents:v1.1:loop_truncated_tools_agent; retrieved 2026-10-04T14:23:50.120Z	64.30 — gemini-3.8-flash; High 推理 · loop_truncated_tools_agent · apex-agents:v1.1:loop_truncated_tools_agent; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent	65.40 — google/gemini-4-argon; High 推理; retrieved 2026-10-04T14:23:50.120Z	61.44 — google/gemini-3.8-flash; High 推理; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing	1518.83 — gemini-4-argon-high; High 推理; retrieved 2026-10-04T14:23:50.120Z	1484.53 — gemini-3.8-flash-high; High 推理; retrieved 2026-10-04T14:23:50.120Z

—

68.90 ±0.97

—

Announced 2026-09-30 in Google's Gemini 4 Argon launch post (https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-4-argon/). Google describes an initial rollout to trusted cyber defenders through the Fairwind Program; broader developer and consumer availability was still forthcoming. This announcement date does not imply general API availability. Not listed in the public ChinaAPI catalogue read on 2026-10-01.

—

MiMo-V2.6-Flash

Xiaomi

China

2026-09-22

yes

—

Unranked: insufficient families, operators, categories. 4 sources, 2 families, 2 operators; 19.0% budget coverage.

Source observations

Arena Text Style-Controlled: 1451.7828617232694 — mimo-v2.6-flash; source_default:default@mimo-v2-6-flash; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 56.277 — xiaomi/mimo-v2.6-flash; source_default:default@xiaomi-mimo-v2-6-flash; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1385.234232012161 — mimo-v2.6-flash; source_default:default@mimo-v2-6-flash; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1247.4693835962391 — mimo-v2.6-flash; source_default:default@mimo-v2-6-flash; retrieved 2026-10-04T14:23:50.120Z

Compare an earlier model: MiMo-V2.5

Same evaluation protocol only; configurations are stated for each entry. No overall rank is inferred.

Source	MiMo-V2.6-Flash	MiMo-V2.5
Arena Text Style-Controlled	1451.78 — mimo-v2.6-flash; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1433.61 — mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent	56.28 — xiaomi/mimo-v2.6-flash; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	36.73 — xiaomi/mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing	1385.23 — mimo-v2.6-flash; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1394.37 — mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled	1247.47 — mimo-v2.6-flash; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1233.18 — mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z

1452 ±8

—

53.23 ±1.10

—

1502 ±27

1521 ±15

—

Released and open-sourced 2026-09-22 per Xiaomi's release record (MiMo-V2.6 Series Release). The 2026-10-01 Vals Index v2.1 read independently covers this model, replacing the previously quoted vendor self-reported DeepSWE figure.

—

MiMo-V2.6-Pro

Xiaomi

China

2026-09-22

yes

—

Unranked: insufficient families. 6 sources, 2 families, 3 operators; 25.4% budget coverage.

Source observations

Arena Text Style-Controlled: 1480.075697959219 — mimo-v2.6-pro; source_default:default@mimo-v2-6-pro; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1618.4140075104192 — mimo-v2.6-pro; source_default:default@mimo-v2-6-pro; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 59.5 — mimo-v2.6-pro; scaffolded:harness-loop-truncated-tools-agent+systemid-apex-agents-v1-1-loop-truncated-tools-agent@mimo-v2-6-pro; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 57.339 — xiaomi/mimo-v2.6-pro; source_default:default@xiaomi-mimo-v2-6-pro; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1446.0788814209818 — mimo-v2.6-pro; source_default:default@mimo-v2-6-pro; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1245.2163080026733 — mimo-v2.6-pro; source_default:default@mimo-v2-6-pro; retrieved 2026-10-04T14:23:50.120Z

Compare an earlier model: MiMo-V2.5

Same evaluation protocol only; configurations are stated for each entry. No overall rank is inferred.

Source	MiMo-V2.6-Pro	MiMo-V2.5
Arena Text Style-Controlled	1480.08 — mimo-v2.6-pro; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1433.61 — mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev	1618.41 — mimo-v2.6-pro; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1437.57 — mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent	57.34 — xiaomi/mimo-v2.6-pro; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	36.73 — xiaomi/mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing	1446.08 — mimo-v2.6-pro; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1394.37 — mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled	1245.22 — mimo-v2.6-pro; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1233.18 — mimo-v2.5; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z

1480 ±9

—

55.20 ±1.18

—

1511 ±33

1540 ±18

—

Released and open-sourced 2026-09-22 per Xiaomi's release record (MiMo-V2.6 Series Release).

—

MiMo-V2.6-Pro-Ultraspeed

Xiaomi

China

2026-09-22

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

Xiaomi's high-speed serving mode of MiMo-V2.6-Pro, released 2026-09-22. The vendor states flagship V2.6-Pro performance at up to 20x the output speed but publishes no separate figures, and no board lists it separately, so MiMo-V2.6-Pro's scores are not reused here.

—

Step 5 Preview

StepFun

China

2026-09-20

yes

—

Unranked: insufficient families, operators. 5 sources, 2 families, 2 operators; 21.4% budget coverage.

Source observations

Arena Text Style-Controlled: 1453.6796377727962 — Step 5 Preview; source_default:default@step-5-preview; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1570.151007082731 — step-5-preview-high; source_default:default@step-5-preview-high; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 50.671 — stepfun/step-5-preview; source_default:default@stepfun-step-5-preview; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1420.904169644511 — Step 5 Preview; source_default:default@step-5-preview; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1259.7345848754278 — Step 5 Preview; source_default:default@step-5-preview; retrieved 2026-10-04T14:23:50.120Z

Compare an earlier model: Step 3.5 Flash

Same evaluation protocol only; configurations are stated for each entry. No overall rank is inferred.

Source	Step 5 Preview	Step 3.5 Flash
Arena Text Style-Controlled	1453.68 — Step 5 Preview; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1393.27 — step-3.5-flash; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing	1420.90 — Step 5 Preview; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z	1345.43 — step-3.5-flash; 来源默认配置; retrieved 2026-10-04T14:23:50.120Z

—

49.35 ±1.22

—

16.7

StepFun's announcement page prints no date; it says the model is released today and cites Artificial Analysis results published as of 2026-09-20, and reports dated that day already quote the release, which fixes the release date. The vendor states open weights follow on 2026-10-15. Vals lists it as Step 5 Preview, released 2026-09-28 per its model page; the release date above follows StepFun's announcement.

ChinaAPI Consensus v15

Claude Opus 5.5

Anthropic

International

2026-09-22

yes

81.8

15 sources, 8 families, 6 operators; 47.0% budget coverage. Sensitivity rank range 1–3 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1503.695568013531 — claude-opus-5.5-high; first_party:high@claude-opus-5-5-high; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1815.3614916224676 — claude-opus-5.5-max; first_party:max@claude-opus-5-5-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 73.5 — claude-opus-5-5; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@claude-opus-5-5; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 58.587 — anthropic/claude-opus-5-5; first_party:max@anthropic-claude-opus-5-5; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 80.485 — claude-opus-5-5-max-effort; first_party:max-effort@claude-opus-5-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 94.615625 — claude-opus-5-5-max-effort; first_party:max-effort@claude-opus-5-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 76.003875 — claude-opus-5-5-max-effort; first_party:max-effort@claude-opus-5-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1516.3121626892585 — claude-opus-5.5-high; first_party:high@claude-opus-5-5-high; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.722 — claude-opus-5-5_max; first_party:max@claude-opus-5-5-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9059343434343434 — claude-opus-5-5_max; first_party:max@claude-opus-5-5-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.9122807017543859 — claude-opus-5-5_max; first_party:max@claude-opus-5-5-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.95 — claude-opus-5-5_max; first_party:max@claude-opus-5-5-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.71 — claude-opus-5-5_max; first_party:max@claude-opus-5-5-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 2050.1 — claude-opus-5-5; source_default:default@claude-opus-5-5; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 84.6 — claude-opus-5-5; source_default:default@claude-opus-5-5; retrieved 2026-10-04T14:23:50.120Z

1504 ±9

83.2

—

167 (164 - 172)

66.97 ±0.89

—

1571 ±29

1538 ±18

89.3

71.7

—

Released 2026-09-22 per Anthropic's announcement (Introducing Claude Opus 5.5), the first model in the Claude 5.5 family. Artificial Analysis printed no output speed for its Claude Opus 5.5 (max with fallback) row on 2026-09-22 (the column reads --), so no speed figure is quoted; the speeds of its lower-effort rows are not substituted.

Claude Fable 5.1

Anthropic

International

2026-09-01

yes

79.5

17 sources, 8 families, 7 operators; 56.4% budget coverage. Sensitivity rank range 2–4 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1501.1356497681938 — claude-fable-5.1-max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1749.2435972083101 — claude-fable-5.1-max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 68.6 — claude-fable-5.1; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@claude-fable-5-1; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 58.877 — anthropic/claude-fable-5-1; first_party:max@anthropic-claude-fable-5-1; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.9048 — claude-fable-5-1; first_party:high@claude-fable-5-1; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 76.21783333333335 — claude-fable-5-1-max-effort; first_party:max-effort@claude-fable-5-1-max-effort; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 94.3495 — claude-fable-5-1-max-effort; first_party:max-effort@claude-fable-5-1-max-effort; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 81.24437499999999 — claude-fable-5-1-max-effort; first_party:max-effort@claude-fable-5-1-max-effort; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1481.7727852039911 — claude-fable-5.1-max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1287.6910785206005 — claude-fable-5.1-max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.708 — claude-fable-5-1_max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.9017543859649123 — claude-fable-5-1_max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.8780487804878049 — claude-fable-5-1_max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.47 — claude-fable-5-1_max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.58 — claude-fable-5-1_max; first_party:max@claude-fable-5-1-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 2162 — claude-fable-5-1; source_default:default@claude-fable-5-1; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 85.3 — claude-fable-5-1; source_default:default@claude-fable-5-1; retrieved 2026-10-04T14:23:50.120Z

1501 ±6

83.4

—

165 (162 - 169)

65.83 ±1.11

—

1570 ±21

1531 ±13

86.4

66.1

71.1

Released 2026-09-01 per Anthropic's announcement (Introducing Claude Fable 5.1 and Claude Mythos 5.1); on sale here since 2026-09-02. LMArena's row is the max-effort build, standing on 11,800 votes in its 2026-10-02 dataset publish; the same row on the Chinese category stands on 874 votes. Epoch's fit as read on 2026-10-04 scores it 164.82 and Claude Fable 5 162.22.

GPT-6 Astra

OpenAI

International

2026-09-03

yes

79.5

19 sources, 10 families, 8 operators; 62.0% budget coverage. Sensitivity rank range 1–6 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1477.11090776938 — gpt-6-astra-max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1787.6659051450358 — gpt-6-astra-max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 64.7 — gpt-6-astra-max; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 53.54 — openai/gpt-6-astra; first_party:max@openai-gpt-6-astra; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.7323008849557522 — gpt-6-astra; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-gpt-6-astra-max+systemid-deepswe-v1-1-mini-swe-agent@gpt-6-astra; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8889 — gpt-6-astra; first_party:xhigh@gpt-6-astra; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 68.84075 — gpt-6-astra-max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 94.72999999999999 — gpt-6-astra-max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 82.50483333333335 — gpt-6-astra-max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1449.0912821587349 — gpt-6-astra-max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1284.3484664349812 — gpt-6-astra-max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.756 — gpt-6-astra_max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9577020202020202 — gpt-6-astra_max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.9368421052631579 — gpt-6-astra_max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.976 — gpt-6-astra_max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.72 — gpt-6-astra_max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.84 — gpt-6-astra_max; first_party:max@gpt-6-astra-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 2173.3 — gpt-6-astra; source_default:default@gpt-6-astra; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 82.8 — gpt-6-astra; source_default:default@gpt-6-astra; retrieved 2026-10-04T14:23:50.120Z

1477 ±7

82.2

—

167 (163 - 171)

63.13 ±1.19

—

1510 ±22

1543 ±13

80.4

57.3

35.0

Released 2026-09-03 per OpenAI's announcement (GPT-6 Astra: A new generation of intelligence). LMArena added gpt-6-astra-max to its text board on 2026-09-11 per its leaderboard changelog; the row stands on 9,156 votes in its 2026-10-02 dataset publish. The Chinese category carried no row for it as of 2026-09-22; in the 2026-10-02 publish it does, on 674 votes. On sale here since 2026-09-05.

Claude Fable 5

Anthropic

International

2026-06-09

yes

78.8

19 sources, 10 families, 8 operators; 62.0% budget coverage. Sensitivity rank range 4–7 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1504.270990068727 — claude-fable-5-high; first_party:high@claude-fable-5-high; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1625.2989118244407 — claude-fable-5-high; first_party:high@claude-fable-5-high; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 63.6 — claude-fable-5; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@claude-fable-5; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 56.314 — anthropic/claude-fable-5; first_party:max@anthropic-claude-fable-5; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.6972477064220184 — claude-fable-5; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-claude-fable-5-max+systemid-deepswe-v1-1-mini-swe-agent@claude-fable-5; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8902 — claude-fable-5; first_party:high@claude-fable-5; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 74.08183333333334 — claude-fable-5-max-effort; first_party:max-effort@claude-fable-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 92.81975 — claude-fable-5-max-effort; first_party:max-effort@claude-fable-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 83.227375 — claude-fable-5-max-effort; first_party:max-effort@claude-fable-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1502.572777815691 — claude-fable-5-high; first_party:high@claude-fable-5-high; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1308.7511219820094 — claude-fable-5-high; first_party:high@claude-fable-5-high; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.707 — claude-fable-5_xhigh; first_party:xhigh@claude-fable-5-xhigh; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.8585858585858586 — claude-fable-5_max; first_party:max@claude-fable-5-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.8701754385964913 — claude-fable-5_max; first_party:max@claude-fable-5-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.902 — claude-fable-5_max; first_party:max@claude-fable-5-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.41 — claude-fable-5_max; first_party:max@claude-fable-5-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.52 — claude-fable-5_max; first_party:max@claude-fable-5-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1943.1 — claude-fable-5; source_default:default@claude-fable-5; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 83 — claude-fable-5; source_default:default@claude-fable-5; retrieved 2026-10-04T14:23:50.120Z

1504 ±4

83.0

—

162 (160 - 166)

61.39 ±1.00

—

1555 ±12

1552 ±7

86.0

62.2

64.4

Artificial Analysis marked its Claude Fable 5 (with fallback) row deprecated between the 2026-09-12 and 2026-09-21 reads; the leaderboard hides such rows under its default Status: Current filter, so this row's leaderboard figures are read with the Status filter set to All. LMArena relabelled this row from claude-fable-5 to claude-fable-5-high between 2026-09-06 and 2026-09-18: the Internet Archive's copies of the board show the same model key (claude-fable-5-text) with its vote count running on from 27,189 to 30,057, so the entry follows the new label on all three LMArena text boards.

Gemini 3.7 Flash

Google

International

2026-08-13

yes

73.6

17 sources, 9 families, 7 operators; 56.0% budget coverage. Sensitivity rank range 5–11 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1487.6771302275815 — gemini-3.7-flash-high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1591.7198842613222 — gemini-3.7-flash-high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 67.8 — gemini-3.7-flash; scaffolded:harness-loop-truncated-tools-agent+high+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gemini-3-7-flash; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 59.042 — google/gemini-3.7-flash; first_party:high@google-gemini-3-7-flash; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.6526548672566371 — gemini-3-7-flash; scaffolded:harness-mini-swe-agent+high+mini-swe-agent-gemini-3-7-flash-high+systemid-deepswe-v1-1-mini-swe-a@gemini-3-7-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 68.58574999999999 — gemini-3.7-flash-high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 90.633 — gemini-3.7-flash-high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 82.690125 — gemini-3.7-flash-high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1493.3628664403816 — gemini-3.7-flash-high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1296.0089955863393 — gemini-3.7-flash-high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.692 — gemini-3.7-flash_high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9482323232323232 — gemini-3.7-flash_high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.7157894736842105 — gemini-3.7-flash_high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.36585365853658536 — gemini-3.7-flash_high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.47 — gemini-3.7-flash_high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.37 — gemini-3.7-flash_high; first_party:high@gemini-3-7-flash-high; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1723.2 — gemini-3.7-flash; source_default:default@gemini-3-7-flash; retrieved 2026-10-04T14:23:50.120Z

1488 ±5

78.8

—

157 (155 - 160)

51.27 ±1.10

—

1546 ±15

1518 ±8

78.9

58.3

281

15.0

Released 2026-08-13. LMArena's row stands on 21,539 votes in its 2026-10-02 dataset publish. Artificial Analysis marked its Gemini 3.7 Flash (high) row deprecated between the 2026-09-12 and 2026-09-21 reads; the leaderboard hides such rows under its default Status: Current filter, so this row's leaderboard figures are read with the Status filter set to All.

GPT-6.1 Sol

OpenAI

International

2026-09-29

yes

73.3

15 sources, 7 families, 5 operators; 48.8% budget coverage. Sensitivity rank range 2–8 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1483.2124603186985 — gpt-6.1-sol-max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1757.8125808243285 — gpt-6.1-sol-max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 60 — gpt-6-1-sol; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gpt-6-1-sol; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 52.033 — openai/gpt-6.1-sol; first_party:max@openai-gpt-6-1-sol; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 67.45191666666666 — gpt-6.1-sol-max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 94.733125 — gpt-6.1-sol-max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 82.140125 — gpt-6.1-sol-max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1459.7198658172026 — gpt-6.1-sol-max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1290.6118265721768 — gpt-6.1-sol-max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.739 — gpt-6.1-sol_max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9539141414141414 — gpt-6.1-sol_max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.9368421052631579 — gpt-6.1-sol_max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 1 — gpt-6.1-sol_max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.61 — gpt-6.1-sol_max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.8 — gpt-6.1-sol_max; first_party:max@gpt-6-1-sol-max; retrieved 2026-10-04T14:23:50.120Z

—

81.6

—

61.15 ±1.01

—

80.4

54.5

—

Released 2026-09-29 per OpenAI's announcement (https://openai.com/index/introducing-gpt-6-1-sol/). This is a distinct model from GPT-6 Sol, whose row and scores are retained separately. LiveBench's Max Effort row is used consistently for Overall, Coding and Agentic Coding. On sale here since 2026-10-07.

Claude Opus 5

Anthropic

International

2026-07-24

yes

72.0

19 sources, 10 families, 8 operators; 62.0% budget coverage. Sensitivity rank range 3–8 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1489.4940018458387 — claude-opus-5-max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1694.979974655013 — claude-opus-5-max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 65.8 — claude-opus-5; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@claude-opus-5; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 58.633 — anthropic/claude-opus-5; first_party:max@anthropic-claude-opus-5; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.7364864864864865 — claude-opus-5; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-claude-opus-5-max+systemid-deepswe-v1-1-mini-swe-agent@claude-opus-5; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8889 — claude-opus-5; first_party:xhigh@claude-opus-5; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 73.32374999999999 — claude-opus-5-max-effort; first_party:max-effort@claude-opus-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 93.47125 — claude-opus-5-max-effort; first_party:max-effort@claude-opus-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 76.22725 — claude-opus-5-max-effort; first_party:max-effort@claude-opus-5-max-effort; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1470.5582170024918 — claude-opus-5-max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1288.999501670278 — claude-opus-5-high; first_party:high@claude-opus-5-high; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.599 — claude-opus-5_max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9387626262626263 — claude-opus-5_max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.856140350877193 — claude-opus-5_max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.7317073170731707 — claude-opus-5_max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.42 — claude-opus-5_max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.59 — claude-opus-5_max; first_party:max@claude-opus-5-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 2132.6 — claude-opus-5; source_default:default@claude-opus-5; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 86.3 — claude-opus-5; source_default:default@claude-opus-5; retrieved 2026-10-04T14:23:50.120Z

1490 ±4

80.1

—

163 (160 - 167)

63.67 ±0.96

—

1548 ±10

1533 ±6

81.4

65.2

55.6

Muse Spark 1.3

Meta

International

2026-08-05

yes

58.4

13 sources, 6 families, 7 operators; 48.6% budget coverage. Sensitivity rank range 11–19 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1493.5119101045673 — muse-spark-1.2 (xHigh); first_party:xhigh@muse-spark-1-2-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1531.7785497869363 — muse-spark-1.2 (xHigh); first_party:xhigh@muse-spark-1-2-xhigh; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 36.4 — muse-spark-1-2; scaffolded:harness-loop-truncated-tools-agent+systemid-apex-agents-v1-1-loop-truncated-tools-agent+xhigh@muse-spark-1-2; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 60.599 — meta/muse_spark_1_2; first_party:xhigh@meta-muse-spark-1-2; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.5486725663716814 — muse-spark-1-2; scaffolded:harness-mini-swe-agent+mini-swe-agent-muse-spark-1-2-xhigh+systemid-deepswe-v1-1-mini-swe-agent+x@muse-spark-1-2; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 67.55858333333333 — muse-spark-1.2-xhigh; first_party:xhigh@muse-spark-1-2-xhigh; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 90.604125 — muse-spark-1.2-xhigh; first_party:xhigh@muse-spark-1-2-xhigh; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 76.44983333333333 — muse-spark-1.2-xhigh; first_party:xhigh@muse-spark-1-2-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1455.4876719827569 — muse-spark-1.2 (xHigh); first_party:xhigh@muse-spark-1-2-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1292.7964563760431 — muse-spark-1.2 (xHigh); first_party:xhigh@muse-spark-1-2-xhigh; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.6030150753768844 — muse-spark-1.2_xhigh; first_party:xhigh@muse-spark-1-2-xhigh; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1840.4 — muse-spark-1.2; source_default:default@muse-spark-1-2; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 81.5 — muse-spark-1.2; source_default:default@muse-spark-1-2; retrieved 2026-10-04T14:23:50.120Z

1494 ±9

78.0

—

155 (153 - 157)

49.29 ±1.10

—

1528 ±31

1531 ±17

77.5

57.6

172

31.1

Epoch added a Muse Spark 1.2 row (dated 2026-08-05) between the 2026-08-31 and 2026-09-05 snapshots. Artificial Analysis marked its Muse Spark 1.2 (xhigh) row deprecated between the 2026-09-12 and 2026-09-21 reads; the leaderboard hides such rows under its default Status: Current filter, so this row's leaderboard figures are read with the Status filter set to All. Served here as muse-spark-1.2-contributor, Meta's Contributor tier, whose terms let Meta train future models on the prompts and completions sent to it; the xhigh setting the LMArena, LiveBench and Artificial Analysis rows quote is available on that tier.

Grok 4.6

xAI

International

2026-08-12

57.0

17 sources, 9 families, 7 operators; 56.0% budget coverage. Sensitivity rank range 14–24 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1453.8262540751045 — grok-4.6-high; first_party:high@grok-4-6-high; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1619.5331014852447 — grok-4.6-high; first_party:high@grok-4-6-high; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 65.3 — grok-4-6-xhigh; scaffolded:harness-loop-truncated-tools-agent+systemid-apex-agents-v1-1-loop-truncated-tools-agent+xhigh@grok-4-6-xhigh; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 53.682 — grok/grok-4.6; first_party:high@grok-grok-4-6; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.6674057649667405 — grok-4-6; scaffolded:harness-mini-swe-agent+mini-swe-agent-grok-4-6-xhigh+systemid-deepswe-v1-1-mini-swe-agent+xhigh@grok-4-6; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.873 — grok-4.6@xhigh; first_party:xhigh@grok-4-6-xhigh; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 66.89800000000001 — grok-4.6; source_default:default@grok-4-6; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 91.53875 — grok-4.6; source_default:default@grok-4-6; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 77.78175 — grok-4.6; source_default:default@grok-4-6; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1444.7775949996321 — grok-4.6-high; first_party:high@grok-4-6-high; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1263.466282985066 — grok-4.6-high; first_party:high@grok-4-6-high; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.489 — grok-4.6_xhigh; first_party:xhigh@grok-4-6-xhigh; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9318181818181818 — grok-4.6_xhigh; first_party:xhigh@grok-4-6-xhigh; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.6596491228070176 — grok-4.6_xhigh; first_party:xhigh@grok-4-6-xhigh; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.3170731707317073 — grok-4.6_xhigh; first_party:xhigh@grok-4-6-xhigh; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.31 — grok-4.6_xhigh; first_party:xhigh@grok-4-6-xhigh; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.34 — grok-4.6_xhigh; first_party:xhigh@grok-4-6-xhigh; retrieved 2026-10-04T14:23:50.120Z

1454 ±5

78.0

—

157 (155 - 159)

52.09 ±1.15

—

1503 ±14

1507 ±8

76.8

57.0

12.2

Released 2026-08-12. LMArena labels this vendor SpaceXAI. Its row stands on 23,680 votes in LMArena's 2026-10-02 dataset publish.

Qwen3.8-Max-0902

Alibaba

China

2026-09-02

yes

56.9

7 sources, 6 families, 3 operators; 18.6% budget coverage. Sensitivity rank range 14–19 across 66 recomputation scenarios (not a confidence interval). Not eligible for a rank in 5 of them.

Source observations

Arena WebDev: 1670.011689642309 — qwen3.8-max-0902; source_default:default@qwen3-8-max-0902; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8333 — qwen3.8-max-0902; first_party:xhigh@qwen3-8-max-0902; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.473 — qwen3.8-max-0902_xhigh; first_party:xhigh@qwen3-8-max-0902-xhigh; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.922979797979798 — qwen3.8-max-0902_xhigh; first_party:xhigh@qwen3-8-max-0902-xhigh; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.656140350877193 — qwen3.8-max-0902_xhigh; first_party:xhigh@qwen3-8-max-0902-xhigh; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.34146341463414637 — qwen3.8-max-0902_xhigh; first_party:xhigh@qwen3-8-max-0902-xhigh; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.4 — qwen3.8-max-0902_xhigh; first_party:xhigh@qwen3-8-max-0902-xhigh; retrieved 2026-10-04T14:23:50.120Z

—

72.62

155 (153 - 157)

—

20.0

Alibaba's 2026-09-02 snapshot of Qwen3.8-Max (alias qwen3.8-max-2026-09-02), listed on Model Studio on 2026-09-02 as an upgraded snapshot of qwen3.8-max; Epoch dates it 2026-09-01. Alibaba Cloud's update notices of 2026-09-02 (www.aliyun.com/notice/118616 and the international Model Studio notice Update Notice for Qwen3.8-Max Models) state that from 10:00 Beijing time on 2026-09-05, subject to the actual change time, the qwen3.8-max model name automatically moved to this snapshot with billing unchanged. Alibaba published no completion notice; Artificial Analysis has since marked the earlier build deprecated. This row therefore links to qwen3.8-max, the name on sale here; the same snapshot is also sold under its pinned name qwen3.8-max-0902. Artificial Analysis and Epoch list it as Qwen3.8 Max (0902), SuperCLUE as Qwen3.8-Max-0902(max) (evaluation published 2026-09-10). LMArena added qwen3.8-max-0902 only to its Code Arena (changelog, 2026-09-01), which this page does not cite, and neither LiveBench nor Vals carries a 0902 row; the 2026-08-02 build's figures on the qwen3.8-max row are not reused here.

GLM-5.3

Zhipu AI (Z.ai)

China

2026-08-14

yes

55.4

18 sources, 10 families, 8 operators; 56.0% budget coverage. Sensitivity rank range 14–20 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1478.49856297707 — glm-5.3-max; source_default:default@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1622.9514135470026 — glm-5.3-max; source_default:default@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 56.6 — glm-5-3; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@glm-5-3; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 55.84 — zai/glm-5.3; first_party:max@zai-glm-5-3; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.6895787139689579 — glm-5-3; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-glm-5-3-max+systemid-deepswe-v1-1-mini-swe-agent@glm-5-3; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.836 — glm-5.3; first_party:max@glm-5-3; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 69.9295 — glm-5.3; source_default:default@glm-5-3; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 86.8505 — glm-5.3; source_default:default@glm-5-3; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 74.58016666666666 — glm-5.3; source_default:default@glm-5-3; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1453.9481714839558 — glm-5.3-max; source_default:default@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.41 — glm-5.3_max; first_party:max@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9090909090909091 — glm-5.3_max; first_party:max@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.6877192982456141 — glm-5.3_max; first_party:max@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.2926829268292683 — glm-5.3_max; first_party:max@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.21 — glm-5.3_max; first_party:max@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.33 — glm-5.3_max; first_party:max@glm-5-3-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 2075 — GLM-5.3; source_default:default@glm-5-3; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 81.8 — GLM-5.3; source_default:default@glm-5-3; retrieved 2026-10-04T14:23:50.120Z

1478 ±6

76.1

71.29

156 (154 - 158)

53.51 ±1.30

—

1531 ±17

1521 ±9

79.0

60.9

48.3

Released 2026-08-14; served here via Zhipu's official channels since 2026-08-18. Independent boards picked it up between the 2026-08-19 and 2026-08-26 snapshots, so the vendor's own DeepSWE figure it carried before has been replaced by their readings; the LMArena row stands on 17,857 votes in its 2026-10-02 dataset publish. Epoch, which carried no row for it as of 2026-09-05, lists it as GLM-5.3 (dated 2026-08-14) in the 2026-09-22 read, and SuperCLUE lists GLM-5.3(max) with an evaluation published 2026-09-10. Vals lists it as GLM 5.3, released 2026-08-18 per its model page, the date of Z.ai's release notes.

Qwen3.8-Flash

Alibaba

China

2026-08-26

yes

54.6

5 sources, 3 families, 3 operators; 15.6% budget coverage. Sensitivity rank range 19–27 across 66 recomputation scenarios (not a confidence interval). Not eligible for a rank in 7 of them.

Source observations

Arena WebDev: 1637.4947899302167 — qwen3.8-flash-next; source_default:default@qwen3-8-flash-next; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8254 — qwen3.8-flash; first_party:xhigh@qwen3-8-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 67.08341666666666 — qwen3.8-flash-next; source_default:default@qwen3-8-flash-next; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 86.600375 — qwen3.8-flash-next; source_default:default@qwen3-8-flash-next; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 75.87575 — qwen3.8-flash-next; source_default:default@qwen3-8-flash-next; retrieved 2026-10-04T14:23:50.120Z

—

76.2

66.41

—

72.6

61.6

—

The boards list this model under its open-weight release name, Qwen3.8-Flash-Next (released 2026-08-26 per the Qwen team). The vendor's announcement states that the production version, with a 1M default context and built-in tools, is served under the name Qwen3.8-Flash, which is the qwen3.8-flash API on sale here since 2026-09-04; the two rows are joined on that statement, not on the row name. Artificial Analysis now prints 40 without an estimated marker (read 2026-09-21). SuperCLUE lists it under the API name itself, Qwen3.8-Flash(max), evaluated through the API with an evaluation published 2026-09-10, so that figure needs no join. Neither Epoch, LMArena nor Vals carried a row as of 2026-09-05.

Gemini 3.6 Flash

Google

International

2026-07-21

yes

54.5

17 sources, 9 families, 7 operators; 56.0% budget coverage. Sensitivity rank range 14–23 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1483.0122992912945 — gemini-3.6-flash-high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1536.2711207883613 — gemini-3.6-flash-high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 46.9 — gemini-3.6-flash; scaffolded:harness-loop-truncated-tools-agent+high+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gemini-3-6-flash; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 56.296 — google/gemini-3.6-flash; first_party:high@google-gemini-3-6-flash; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.4668141592920354 — gemini-3-6-flash; scaffolded:harness-mini-swe-agent+high+mini-swe-agent-gemini-3-6-flash-high+systemid-deepswe-v1-1-mini-swe-a@gemini-3-6-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 60.64866666666667 — gemini-3.6-flash-high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 85.77575 — gemini-3.6-flash-high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 79.63220833333332 — gemini-3.6-flash-high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1470.8889674752343 — gemini-3.6-flash-high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1279.6882412124646 — gemini-3.6-flash-high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.662 — gemini-3.6-flash_high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9412878787878788 — gemini-3.6-flash_high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.5894736842105263 — gemini-3.6-flash_high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.21951219512195122 — gemini-3.6-flash_high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.4 — gemini-3.6-flash_high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.3 — gemini-3.6-flash_high; first_party:high@gemini-3-6-flash-high; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1603.5 — gemini-3.6-flash; source_default:default@gemini-3-6-flash; retrieved 2026-10-04T14:23:50.120Z

1483 ±4

73.6

—

154 (153 - 156)

—

1530 ±12

1519 ±7

77.9

43.4

180

14.4

Artificial Analysis marked its Gemini 3.6 Flash row deprecated between the 2026-09-12 and 2026-09-21 reads; the leaderboard hides such rows under its default Status: Current filter, so this row's leaderboard figures are read with the Status filter set to All. The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

DeepSeek V4.1 Flash

DeepSeek

China

2026-09-10

yes

51.6

11 sources, 5 families, 6 operators; 38.6% budget coverage. Sensitivity rank range 20–27 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1474.4402781653991 — deepseek-v4.1-flash-max; source_default:default@deepseek-v4-1-flash-max; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1619.9147841477075 — deepseek-v4.1-flash-max; source_default:default@deepseek-v4-1-flash-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 39.5 — deepseek-v4-1-flash; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@deepseek-v4-1-flash; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 53.481 — deepseek/deepseek-v4.1-flash; first_party:high@deepseek-deepseek-v4-1-flash; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8254 — deepseek-v4.1-flash; first_party:max@deepseek-v4-1-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 78.65483333333333 — deepseek-v4.1-flash-max; source_default:default@deepseek-v4-1-flash-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 89.991 — deepseek-v4.1-flash-max; source_default:default@deepseek-v4-1-flash-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 75.61145833333333 — deepseek-v4.1-flash-max; source_default:default@deepseek-v4-1-flash-max; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1440.14353125847 — deepseek-v4.1-flash-max; source_default:default@deepseek-v4-1-flash-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1540.1 — deepseek-v4.1-flash; source_default:default@deepseek-v4-1-flash; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 74 — deepseek-v4.1-flash; source_default:default@deepseek-v4-1-flash; retrieved 2026-10-04T14:23:50.120Z

1474 ±7

81.1

71.81

155 (149 - 158)

51.32 ±1.13

—

1523 ±23

1528 ±12

80.0

77.3

220

22.8

Released 2026-09-10 per DeepSeek's official announcement and served by the canonical deepseek-flash API model. The vendor retired V4 Flash and V4 Flash Vision Exp that day and temporarily routes their old identifiers to V4.1 Flash; their historical leaderboard rows remain separate and none of their figures are reused here. Artificial Analysis, LiveBench and Vals all published distinct V4.1 Flash rows by 2026-09-12; Epoch (DeepSeek V4.1 Flash, dated 2026-09-09) and SuperCLUE (DeepSeek-V4.1-Flash(max), evaluation published 2026-09-11) carry one as of the 2026-09-22 read.

GPT-5.6 Terra

OpenAI

International

2026-07-09

yes

51.1

19 sources, 10 families, 8 operators; 62.0% budget coverage. Sensitivity rank range 15–24 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1465.4756435259505 — gpt-5.6-terra-xhigh; first_party:xhigh@gpt-5-6-terra-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1519.0885783683277 — gpt-5.6-terra-xhigh (codex-harness); scaffolded:name-codex-harness+xhigh@gpt-5-6-terra-xhigh-codex-harness; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 58.2 — gpt-5.6-terra-max; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 54.436 — openai/gpt-5.6-terra; first_party:max@openai-gpt-5-6-terra; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.6962305986696231 — gpt-5-6-terra; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-gpt-5-6-terra-max+systemid-deepswe-v1-1-mini-swe-agent@gpt-5-6-terra; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8175 — gpt-5.6-terra; first_party:xhigh@gpt-5-6-terra; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 66.59758333333333 — gpt-5.6-terra-max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 92.771375 — gpt-5.6-terra-max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 73.75458333333333 — gpt-5.6-terra-max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1425.5053732054403 — gpt-5.6-terra-xhigh; first_party:xhigh@gpt-5-6-terra-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1268.213209933113 — gpt-5.6-terra-xhigh; first_party:xhigh@gpt-5-6-terra-xhigh; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.432 — gpt-5.6-terra_max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9330808080808081 — gpt-5.6-terra_max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.8596491228070176 — gpt-5.6-terra_max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.7073170731707317 — gpt-5.6-terra_max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.54 — gpt-5.6-terra_max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.35 — gpt-5.6-terra_max; first_party:max@gpt-5-6-terra-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1854.9 — gpt-5.6-terra; source_default:default@gpt-5-6-terra; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 78 — gpt-5.6-terra; source_default:default@gpt-5-6-terra; retrieved 2026-10-04T14:23:50.120Z

1465 ±4

77.9

—

160 (157 - 163)

53.09 ±1.29

—

1516 ±12

1515 ±7

78.2

54.9

31.7

DeepSeek V4 Pro 0813

DeepSeek

China

2026-08-13

yes

49.2

15 sources, 8 families, 6 operators; 46.4% budget coverage. Sensitivity rank range 21–28 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1464.278528826575 — deepseek-v4-pro-high-20260813; source_default:default@deepseek-v4-pro-high-20260813; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1582.6387775192716 — deepseek-v4-pro-high-20260813; source_default:default@deepseek-v4-pro-high-20260813; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 47.3 — deepseek-v4-pro-08-13; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@deepseek-v4-pro-08-13; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 50.393 — deepseek/deepseek-v4-pro-0813; first_party:max@deepseek-deepseek-v4-pro-0813; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.7778 — deepseek-v4-pro; source_default:default@deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 66.05408333333334 — deepseek-v4-pro-0813; source_default:default@deepseek-v4-pro-0813; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 90.46375 — deepseek-v4-pro-0813; source_default:default@deepseek-v4-pro-0813; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 74.887 — deepseek-v4-pro-0813; source_default:default@deepseek-v4-pro-0813; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1444.5048285470523 — deepseek-v4-pro-high-20260813; source_default:default@deepseek-v4-pro-high-20260813; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.5290581162324649 — deepseek-v4-pro-0813_max; first_party:max@deepseek-v4-pro-0813-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9166666666666666 — deepseek-v4-pro-0813_max; first_party:max@deepseek-v4-pro-0813-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.6456140350877193 — deepseek-v4-pro-0813_max; first_party:max@deepseek-v4-pro-0813-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.2682926829268293 — deepseek-v4-pro-0813_max; first_party:max@deepseek-v4-pro-0813-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.47 — deepseek-v4-pro-0813_max; first_party:max@deepseek-v4-pro-0813-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.43 — deepseek-v4-pro-0813_max; first_party:max@deepseek-v4-pro-0813-max; retrieved 2026-10-04T14:23:50.120Z

1464 ±7

77.4

70.10

155 (154 - 158)

47.63 ±1.12

—

1493 ±22

1505 ±11

77.2

54.9

17.8

DeepSeek's change log dates this build's GA release to 2026-08-13 and states that the deepseek-v4-pro model name now serves it. Epoch added a DeepSeek V4 Pro 0813 row between the 2026-08-26 and 2026-08-31 snapshots, so the undated DeepSeek-V4-Pro figure is no longer borrowed for it. LMArena's 0813 row stands on 9,780 votes in its 2026-10-02 dataset publish. SuperCLUE announced on 2026-08-26 that it had added the official DeepSeek-V4-Pro; in the July 2026 edition's data file read on 2026-09-22 that row is labelled DeepSeek-V4-Pro-0813(max), with an evaluation published 2026-08-18, and the 64.40 row the board had labelled DeepSeek-V4-Pro(max) 预览版, now plain DeepSeek-V4-Pro(max), is the preview build, quoted on the deepseek-v4-pro row rather than here.

Grok 4.7

xAI

International

2026-09-21

yes

48.1

16 sources, 8 families, 6 operators; 48.8% budget coverage. Sensitivity rank range 20–26 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1442.4571916368702 — grok-4.7-xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1637.5244425163132 — grok-4.7-xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 54.6 — grok-4.7; scaffolded:harness-loop-truncated-tools-agent+systemid-apex-agents-v1-1-loop-truncated-tools-agent+xhigh@grok-4-7; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 52.251 — grok/grok-4.7; first_party:xhigh@grok-grok-4-7; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 65.57408333333333 — grok-4.7-xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 89.165875 — grok-4.7-xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 77.70695833333333 — grok-4.7-xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1435.6958125452798 — grok-4.7-xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.56 — grok-4.7_xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.92739898989899 — grok-4.7_xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.5298245614035088 — grok-4.7_xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.17073170731707318 — grok-4.7_xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.38 — grok-4.7_xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.29 — grok-4.7_xhigh; first_party:xhigh@grok-4-7-xhigh; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 2006.7 — grok-4.7; source_default:default@grok-4-7; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 83 — grok-4.7; source_default:default@grok-4-7; retrieved 2026-10-04T14:23:50.120Z

1442 ±8

77.4

—

54.95 ±1.07

—

1507 ±25

1488 ±16

77.2

54.0

15.0

Released 2026-09-21 per the vendor's announcement, which now publishes under the SpaceXAI name; the row keeps the xAI label of the earlier Grok rows. Artificial Analysis lists Grok 4.7 (xhigh) and Grok 4.7 (high), both printing 46; the xhigh row is quoted because it is the highest effort setting and the higher score before rounding (46.4 against 46.3).

Grok 4.5

xAI

International

2026-07-08

47.3

17 sources, 10 families, 8 operators; 58.4% budget coverage. Sensitivity rank range 25–29 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1465.785937547922 — grok-4.5; source_default:default@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1551.8211671727047 — grok-4.5; source_default:default@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 56.2 — grok-4-5; scaffolded:harness-loop-truncated-tools-agent+high+systemid-apex-agents-v1-1-loop-truncated-tools-agent@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 48.349 — grok/grok-4.5; first_party:high@grok-grok-4-5; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.5376106194690266 — grok-4-5; scaffolded:harness-mini-swe-agent+high+mini-swe-agent-grok-4-5-high+systemid-deepswe-v1-1-mini-swe-agent@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.7831 — grok-4.5; first_party:high@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 62.52508333333333 — grok-4.5; source_default:default@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 88.99875 — grok-4.5; source_default:default@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 77.16229166666668 — grok-4.5; source_default:default@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1450.2019696971677 — grok-4.5; source_default:default@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1278.980636729535 — grok-4.5; source_default:default@grok-4-5; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.483 — grok-4.5_high; first_party:high@grok-4-5-high; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9343434343434344 — grok-4.5_high; first_party:high@grok-4-5-high; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.5719298245614035 — grok-4.5_high; first_party:high@grok-4-5-high; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.24390243902439024 — grok-4.5_high; first_party:high@grok-4-5-high; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.36 — grok-4.5_high; first_party:high@grok-4-5-high; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1578.8 — grok-4.5; source_default:default@grok-4-5; retrieved 2026-10-04T14:23:50.120Z

1466 ±4

75.8

—

154 (152 - 156)

44.72 ±1.17

—

1504 ±11

1514 ±7

68.6

56.5

15.0

LMArena labels this vendor SpaceXAI. Artificial Analysis marked its Grok 4.5 (high) row deprecated between the 2026-09-12 and 2026-09-21 reads; the leaderboard hides such rows under its default Status: Current filter, so this row's leaderboard figures are read with the Status filter set to All. Released 2026-07-08 per xAI's announcement as published that day and its API release notes; the live announcement page now shows Jul 16, the date it was edited when the model reached the EU.

Gemini 3.1 Pro

Google

International

2026-02-19

46.7

19 sources, 10 families, 8 operators; 62.0% budget coverage. Sensitivity rank range 21–29 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1486.9848912833006 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1446.2003008341799 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 35.3 — gemini-3.1-pro-preview-high; scaffolded:harness-loop-truncated-tools-agent+high+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gemini-3-1-pro-preview-high; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 42.982 — google/gemini-3.1-pro-preview; first_party:high@google-gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.1172566371681416 — gemini-3-1-pro-preview; scaffolded:harness-mini-swe-agent+high+mini-swe-agent-gemini-3-1-pro-preview-high+systemid-deepswe-v1-1-mini@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.647 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 60.297916666666666 — gemini-3.1-pro-preview-high; first_party:high@gemini-3-1-pro-preview-high; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 87.524875 — gemini-3.1-pro-preview-high; first_party:high@gemini-3-1-pro-preview-high; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 82.238 — gemini-3.1-pro-preview-high; first_party:high@gemini-3-1-pro-preview-high; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1480.3283746534428 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1279.3114610010014 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.735 — gemini-3.1-pro-preview_high; first_party:high@gemini-3-1-pro-preview-high; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9444444444444444 — gemini-3.1-pro-preview_high; first_party:high@gemini-3-1-pro-preview-high; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.5964912280701754 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.2682926829268293 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.49 — gemini-3.1-pro-preview_high; first_party:high@gemini-3-1-pro-preview-high; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.34 — gemini-3.1-pro-preview_high; first_party:high@gemini-3-1-pro-preview-high; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1491.3 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 68.2 — gemini-3.1-pro-preview; source_default:default@gemini-3-1-pro-preview; retrieved 2026-10-04T14:23:50.120Z

1487 ±3

77.0

—

155 (153 - 157)

33.44 ±1.13

—

1529 ±7

1522 ±5

76.5

44.1

115

15.6

DeepSeek V4 Flash 0731

DeepSeek

China

2026-07-31

46.5

13 sources, 8 families, 6 operators; 36.4% budget coverage. Sensitivity rank range 26–32 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena WebDev: 1580.982689861258 — deepseek-v4-flash-high; source_default:default@deepseek-v4-flash-high; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 55.3 — deepseek-v4-flash; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@deepseek-v4-flash; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 49.517 — deepseek/deepseek-v4-flash-0731; first_party:high@deepseek-deepseek-v4-flash-0731; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.746 — deepseek-v4-flash; first_party:max@deepseek-v4-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 60.87608333333333 — deepseek-v4-flash-0731; source_default:default@deepseek-v4-flash-0731; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 86.712125 — deepseek-v4-flash-0731; source_default:default@deepseek-v4-flash-0731; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 72.345875 — deepseek-v4-flash-0731; source_default:default@deepseek-v4-flash-0731; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.33633633633633636 — deepseek-v4-flash-0731_max; first_party:max@deepseek-v4-flash-0731-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9103535353535354 — deepseek-v4-flash-0731_max; first_party:max@deepseek-v4-flash-0731-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.5754385964912281 — deepseek-v4-flash-0731_max; first_party:max@deepseek-v4-flash-0731-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.24390243902439024 — deepseek-v4-flash-0731_max; first_party:max@deepseek-v4-flash-0731-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.33 — deepseek-v4-flash-0731_max; first_party:max@deepseek-v4-flash-0731-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.34 — deepseek-v4-flash-0731_max; first_party:max@deepseek-v4-flash-0731-max; retrieved 2026-10-04T14:23:50.120Z

—

74.2

65.60

155 (152 - 157)

47.96 ±1.24

—

75.0

46.8

221

13.3

Historical 2026-07-31 V4 Flash build. DeepSeek retired V4 Flash on 2026-09-10 and its compatibility ID now routes to V4.1 Flash; this historical score row is not the currently served build and is not marked on sale. Artificial Analysis marked its DeepSeek V4 Flash 0731 (max) row deprecated between the 2026-09-12 and 2026-09-21 reads; the leaderboard hides such rows under its default Status: Current filter, so this row's leaderboard figures are read with the Status filter set to All.

Qwen3.7-Max

Alibaba

China

2026-05-20

yes

46.4

12 sources, 7 families, 4 operators; 36.4% budget coverage. Sensitivity rank range 22–30 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1474.882310614849 — qwen3.7-max-preview; source_default:default@qwen3-7-max-preview; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 47.776 — alibaba/qwen3.7-max; source_default:default@alibaba-qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 58.9025 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 84.29237499999999 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 76.89025 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1441.2226404517887 — qwen3.7-max-preview; source_default:default@qwen3-7-max-preview; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.5576730190571715 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9090909090909091 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.6456140350877193 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.34146341463414637 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.19 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.32 — qwen3.7-max; source_default:default@qwen3-7-max; retrieved 2026-10-04T14:23:50.120Z

1475 ±10

73.1

—

154 (152 - 156)

—

1527 ±35

1524 ±18

74.2

43.6

—

LMArena lists the preview build (qwen3.7-max-preview); the released model has no separate row. Released 2026-05-20 (10:00 China time) per the Qwen blog; Epoch records 2026-05-19, and the blog itself renders the date in the reader's time zone, so US readers see the 19th. The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

GLM-5.2

Zhipu AI (Z.ai)

China

2026-06-16

yes

45.1

17 sources, 9 families, 7 operators; 52.4% budget coverage. Sensitivity rank range 25–32 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1475.7694821463967 — glm-5.2-max; source_default:default@glm-5-2-max; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1605.0010787624549 — glm-5.2-max; source_default:default@glm-5-2-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 45.2 — glm-5-2; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@glm-5-2; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 49.699 — zai/glm-5.2; source_default:default@zai-glm-5-2; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.43777777777777777 — glm-5-2; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-glm-5-2-max+systemid-deepswe-v1-1-mini-swe-agent@glm-5-2; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 65.71083333333333 — glm-5.2; source_default:default@glm-5-2; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 84.203125 — glm-5.2; source_default:default@glm-5-2; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 69.26670833333333 — glm-5.2; source_default:default@glm-5-2; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1454.1688463438645 — glm-5.2-max; source_default:default@glm-5-2-max; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.342 — glm-5.2_max; first_party:max@glm-5-2-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9185606060606061 — glm-5.2_max; first_party:max@glm-5-2-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.592057761732852 — glm-5.2_max; first_party:max@glm-5-2-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.2926829268292683 — glm-5.2_max; first_party:max@glm-5-2-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.21 — glm-5.2_max; first_party:max@glm-5-2-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.15 — glm-5.2_medium; first_party:medium@glm-5-2-medium; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1756.5 — zai-org/GLM-5.2; source_default:default@zai-org-glm-5-2; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 77.9 — zai-org/GLM-5.2; source_default:default@zai-org-glm-5-2; retrieved 2026-10-04T14:23:50.120Z

1476 ±4

73.2

63.27

152 (150 - 154)

—

1520 ±11

1512 ±6

79.7

51.8

7.2

Artificial Analysis marked its GLM-5.2 (max) row deprecated between the 2026-09-12 and 2026-09-21 reads; the leaderboard hides such rows under its default Status: Current filter, so this row's leaderboard figures are read with the Status filter set to All. The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

GLM-5.3-Flash

Zhipu AI (Z.ai)

China

2026-08-26

yes

43.7

16 sources, 8 families, 7 operators; 52.0% budget coverage. Sensitivity rank range 23–34 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1473.0004900163892 — glm-5.3-flash; source_default:default@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1615.5658203126495 — glm-5.3-flash; source_default:default@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 52.8 — glm-5-3-flash; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 57.85 — zai/glm-5.3-flash; first_party:max@zai-glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.6339285714285714 — glm-5-3-flash; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-glm-5-3-flash-max+systemid-deepswe-v1-1-mini-swe-agent@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8148 — glm-5.3-flash; first_party:max@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 67.85883333333334 — glm-5.3-flash; source_default:default@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 79.44437500000001 — glm-5.3-flash; source_default:default@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 65.06404166666667 — glm-5.3-flash; source_default:default@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1433.3247774862834 — glm-5.3-flash; source_default:default@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1278.383956913884 — glm-5.3-flash; source_default:default@glm-5-3-flash; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9015151515151515 — glm-5.3-flash_max; first_party:max@glm-5-3-flash-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.5578947368421052 — glm-5.3-flash_max; first_party:max@glm-5-3-flash-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.17073170731707318 — glm-5.3-flash_max; first_party:max@glm-5-3-flash-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.14 — glm-5.3-flash_max; first_party:max@glm-5-3-flash-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.08 — glm-5.3-flash_max; first_party:max@glm-5-3-flash-max; retrieved 2026-10-04T14:23:50.120Z

1473 ±5

71.6

68.10

152 (149 - 154)

—

1529 ±15

1523 ±8

79.0

56.8

51.1

Released 2026-08-26 per Z.ai's own release notes; the first natively multimodal model in the GLM-5 line and its low-cost tier, served here on Zhipu's official channels. LMArena, LiveBench and Artificial Analysis all picked it up within days of release; the LMArena row is a ranked one, standing on 22,971 votes in its 2026-10-02 dataset publish. Epoch, SuperCLUE and Vals carried no row for it as of 2026-08-31. In the 2026-09-22 read Epoch lists it as GLM-5.3-Flash dated 2026-08-20, six days before Z.ai's release notes; the benchmark runs behind that figure are on the glm-5.3-flash model itself (Epoch's model versions glm-5.3-flash_max and glm-5.3-flash_high), so only the date differs. SuperCLUE lists GLM-5.3-Flash(max) with an evaluation published 2026-09-10, and Vals lists it as GLM 5.3 Flash, released 2026-08-26 per its model page. GLM-5.3's figures are not reused here. The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

DeepSeek V4 Pro

DeepSeek

China

2026-04-24

42.9

14 sources, 8 families, 5 operators; 42.4% budget coverage. Sensitivity rank range 30–38 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1457.7609688780526 — deepseek-v4-pro; source_default:default@deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1445.9719348913807 — deepseek-v4-pro; source_default:default@deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 44.083 — deepseek/deepseek-v4-pro; first_party:max@deepseek-deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.6283185840707964 — deepseek-v4-pro; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-deepseek-v4-pro-max+systemid-deepswe-v1-1-mini-swe-agen@deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 56.31 — deepseek-v4-pro; source_default:default@deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 86.683875 — deepseek-v4-pro; source_default:default@deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 70.24016666666667 — deepseek-v4-pro; source_default:default@deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1445.9606463389398 — deepseek-v4-pro; source_default:default@deepseek-v4-pro; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.46993987975951906 — deepseek-v4-pro_max; first_party:max@deepseek-v4-pro-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.8964646464646465 — deepseek-v4-pro_max; first_party:max@deepseek-v4-pro-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.45263157894736844 — deepseek-v4-pro_max; first_party:max@deepseek-v4-pro-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.024390243902439025 — deepseek-v4-pro_max; first_party:max@deepseek-v4-pro-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.2 — deepseek-v4-pro_max; first_party:max@deepseek-v4-pro-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.17 — deepseek-v4-pro_none; first_party:non-reasoning@deepseek-v4-pro-none; retrieved 2026-10-04T14:23:50.120Z

1458 ±4

—

64.40

149 (147 - 151)

38.63 ±1.16

—

1489 ±11

1500 ±6

—

21.1

The pre-0813 build; the deepseek-v4-pro API moved to DeepSeek-V4-Pro-0813 on 2026-08-12. LiveBench dropped this row in favour of the dated builds between the 2026-08-19 and 2026-08-26 snapshots, so its 71.6 is no longer quoted. SuperCLUE's DeepSeek-V4-Pro(max) row (64.40), which its July 2026 edition labelled 预览版, is quoted here from CA-971 on, overturning CA-605. CA-605 left it out because LMArena lists deepseek-v4-pro and deepseek-v4-pro-high-preview separately, but LMArena's own data gives those two rows the same checkpoint key (deepseek-v4-pro-ch1-text and deepseek-v4-pro-ch1-thinking-text) and links both to DeepSeek's 2026-04-24 announcement: they are one build in two modes, which DeepSeek's change log calls V4-Pro-Preview. Before the 2026-08-13 GA that build was the only one the deepseek-v4-pro API served, so an API evaluation published in the July edition can only be this row. SuperCLUE's data file has since dropped the 预览版 label and dates the row 2026.9.10, keeping the July figure. Artificial Analysis marked its DeepSeek V4 Pro (max) row deprecated between the 2026-09-12 and 2026-09-21 reads; the leaderboard hides such rows under its default Status: Current filter, so this row's leaderboard figures are read with the Status filter set to All.

GPT-5.6 Luna

OpenAI

International

2026-07-09

yes

42.5

19 sources, 10 families, 8 operators; 62.0% budget coverage. Sensitivity rank range 22–34 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1453.050254337715 — gpt-5.6-luna-xhigh; first_party:xhigh@gpt-5-6-luna-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1517.97562708065 — gpt-5.6-luna-xhigh (codex-harness); scaffolded:name-codex-harness+xhigh@gpt-5-6-luna-xhigh-codex-harness; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 43 — gpt-5.6-luna-max; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 55.044 — openai/gpt-5.6-luna; first_party:max@openai-gpt-5-6-luna; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.671875 — gpt-5-6-luna; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-gpt-5-6-luna-max+systemid-deepswe-v1-1-mini-swe-agent@gpt-5-6-luna; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.7778 — gpt-5.6-luna; first_party:xhigh@gpt-5-6-luna; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 65.67466666666667 — gpt-5.6-luna-max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 86.4225 — gpt-5.6-luna-max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 66.344 — gpt-5.6-luna-max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1408.999823813769 — gpt-5.6-luna-xhigh; first_party:xhigh@gpt-5-6-luna-xhigh; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1259.811764540452 — gpt-5.6-luna-xhigh; first_party:xhigh@gpt-5-6-luna-xhigh; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.41 — gpt-5.6-luna_max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9160353535353535 — gpt-5.6-luna_max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.8210526315789474 — gpt-5.6-luna_max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.6097560975609756 — gpt-5.6-luna_max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.4 — gpt-5.6-luna_max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.21 — gpt-5.6-luna_max; first_party:max@gpt-5-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1828.7 — gpt-5.6-luna; source_default:default@gpt-5-6-luna; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 75.5 — gpt-5.6-luna; source_default:default@gpt-5-6-luna; retrieved 2026-10-04T14:23:50.120Z

1453 ±4

73.6

—

156 (154 - 159)

51.69 ±1.06

—

1476 ±12

1501 ±7

82.9

48.4

144

19.4

The low-cost tier of the GPT-5.6 line. OpenAI announced a limited partner preview of the GPT-5.6 series (Sol, Terra and Luna) on 2026-06-26 and made the models generally available on 2026-07-09, the post the preview page links as the launch; release_date follows the launch.

GPT-6 Luna

OpenAI

International

2026-09-22

yes

39.3

14 sources, 7 families, 5 operators; 42.8% budget coverage. Sensitivity rank range 29–39 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1442.8801590310995 — gpt-6-luna-max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1578.7133557951634 — gpt-6-luna-max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 44.3 — gpt-6-luna; scaffolded:harness-loop-truncated-tools-agent+max+systemid-apex-agents-v1-1-loop-truncated-tools-agent@gpt-6-luna; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 49.873 — openai/gpt-6-luna; first_party:max@openai-gpt-6-luna; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 65.081 — gpt-6-luna-max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 85.446375 — gpt-6-luna-max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 64.88045833333334 — gpt-6-luna-max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1413.255347455926 — gpt-6-luna-max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.414 — gpt-6-luna_max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9046717171717171 — gpt-6-luna_max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.7894736842105263 — gpt-6-luna_max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.5609756097560976 — gpt-6-luna_max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.31 — gpt-6-luna_max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.07 — gpt-6-luna_max; first_party:max@gpt-6-luna-max; retrieved 2026-10-04T14:23:50.120Z

1443 ±7

72.0

—

51.22 ±1.07

—

1484 ±21

1512 ±12

79.0

51.2

157

—

Released 2026-09-22 per OpenAI's announcement (Introducing GPT-6 Sol and Luna) and the API changelog entry of the same day.

Kimi K2.6

Moonshot AI

China

2026-04-20

yes

38.4

16 sources, 8 families, 5 operators; 50.8% budget coverage. Sensitivity rank range 32–36 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1460.9594397305113 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1508.7407156996771 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 44.899 — kimi/kimi-k2.6; source_default:default@kimi-kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 62.74316666666667 — kimi-k2.6-thinking; source_default:default@kimi-k2-6-thinking; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 81.825625 — kimi-k2.6-thinking; source_default:default@kimi-k2-6-thinking; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 69.74979166666667 — kimi-k2.6-thinking; source_default:default@kimi-k2-6-thinking; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1433.184164684043 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1264.9235591095528 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.349 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9078282828282829 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.5719298245614035 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.2564102564102564 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.26 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.18 — kimi-k2.6; source_default:default@kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1724.5 — moonshotai/Kimi-K2.6; source_default:default@moonshotai-kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 78.5 — moonshotai/Kimi-K2.6; source_default:default@moonshotai-kimi-k2-6; retrieved 2026-10-04T14:23:50.120Z

1461 ±4

70.5

—

151 (149 - 153)

—

1516 ±13

1516 ±7

78.6

46.9

—

The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

GLM-5.1

Zhipu AI (Z.ai)

China

2026-04-07

yes

37.5

11 sources, 7 families, 5 operators; 36.8% budget coverage. Sensitivity rank range 31–37 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1464.565468868514 — glm-5.1; source_default:default@glm-5-1; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1508.495438182888 — glm-5.1; source_default:default@glm-5-1; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 40.9 — glm-5-1; scaffolded:harness-loop-truncated-tools-agent+systemid-apex-agents-v1-1-loop-truncated-tools-agent@glm-5-1; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 44.792 — zai/glm-5.1; source_default:default@zai-glm-5-1; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1446.4268189544591 — glm-5.1; source_default:default@glm-5-1; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.34 — glm-5.1; source_default:default@glm-5-1; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.898989898989899 — glm-5.1; source_default:default@glm-5-1; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.3684210526315789 — glm-5.1; source_default:default@glm-5-1; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.19 — glm-5.1; source_default:default@glm-5-1; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1592.2 — zai-org/GLM-5.1; source_default:default@zai-org-glm-5-1; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 73.5 — zai-org/GLM-5.1; source_default:default@zai-org-glm-5-1; retrieved 2026-10-04T14:23:50.120Z

1465 ±4

—

150 (148 - 152)

—

1514 ±10

1513 ±6

—

The Vals Index V2 (2026-08-13) no longer lists this model; the V1.2 figure it once carried was removed with that revision.

MiMo-V2.5-Pro

Xiaomi

China

2026-04-22

yes

35.8

6 sources, 3 families, 3 operators; 21.4% budget coverage. Sensitivity rank range 33–40 across 66 recomputation scenarios (not a confidence interval). Not eligible for a rank in 5 of them.

Source observations

Arena Text Style-Controlled: 1467.5147388759524 — mimo-v2.5-pro; source_default:default@mimo-v2-5-pro; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1477.6447429129985 — mimo-v2.5-pro; source_default:default@mimo-v2-5-pro; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 41.502 — xiaomi/mimo-v2.5-pro; source_default:default@xiaomi-mimo-v2-5-pro; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1434.015825782703 — mimo-v2.5-pro; source_default:default@mimo-v2-5-pro; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1493.4 — XiaomiMiMo/MiMo-V2.5-Pro; source_default:default@xiaomimimo-mimo-v2-5-pro; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 73.7 — XiaomiMiMo/MiMo-V2.5-Pro; source_default:default@xiaomimimo-mimo-v2-5-pro; retrieved 2026-10-04T14:23:50.120Z

1468 ±4

—

53.01

—

33.84 ±1.17

—

1515 ±10

1522 ±6

—

9.4

GLM-5

Zhipu AI (Z.ai)

China

2026-02-12

yes

35.6

7 sources, 4 families, 3 operators; 22.2% budget coverage. Sensitivity rank range 35–40 across 66 recomputation scenarios (not a confidence interval). Not eligible for a rank in 3 of them.

Source observations

Arena Text Style-Controlled: 1457.6119175864085 — glm-5; source_default:default@glm-5; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1434.041012903606 — glm-5; source_default:default@glm-5; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1449.0176431617106 — glm-5; source_default:default@glm-5; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.8781725888324873 — glm-5; source_default:default@glm-5; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.1 — glm-5; source_default:default@glm-5; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1600.6 — zai-org/GLM-5; source_default:default@zai-org-glm-5; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 70.9 — zai-org/GLM-5; source_default:default@zai-org-glm-5; retrieved 2026-10-04T14:23:50.120Z

1458 ±4

—

146 (144 - 148)

—

1516 ±14

1498 ±7

—

Released 2026-02-12 per Z.ai's blog (GLM-5: From Vibe Coding to Agentic Engineering) and both of its release notes; Epoch records 2026-02-11, the US-time day of the launch post.

Qwen3.7-Plus

Alibaba

China

2026-06-01

yes

34.5

8 sources, 5 families, 3 operators; 27.6% budget coverage. Sensitivity rank range 36–41 across 66 recomputation scenarios (not a confidence interval). Not eligible for a rank in 4 of them.

Source observations

Arena Text Style-Controlled: 1455.774883803081 — qwen3.7-plus; source_default:default@qwen3-7-plus; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 38.22 — alibaba/qwen3.7-plus; source_default:default@alibaba-qwen3-7-plus; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1431.8224321990658 — qwen3.7-plus; source_default:default@qwen3-7-plus; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1265.4888736267915 — qwen3.7-plus; source_default:default@qwen3-7-plus; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.8787878787878788 — qwen3.7-plus; source_default:default@qwen3-7-plus; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.34385964912280703 — qwen3.7-plus_none; first_party:non-reasoning@qwen3-7-plus-none; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.24 — qwen3.7-plus; source_default:default@qwen3-7-plus; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.17 — qwen3.7-plus; source_default:default@qwen3-7-plus; retrieved 2026-10-04T14:23:50.120Z

1456 ±4

—

147 (146 - 149)

—

1504 ±12

1501 ±6

—

9.4

Released 2026-06-01 (10:00 China time) per the Qwen blog, the day Model Studio also lists it; Epoch records 2026-06-02, the China-time day of the launch post on X. The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

Qwen3.8-27B

Alibaba

China

2026-08-14

yes

32.5

12 sources, 5 families, 6 operators; 44.6% budget coverage. Sensitivity rank range 27–40 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1437.6997264231165 — qwen3.8-27b; source_default:default@qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1591.235093957948 — qwen3.8-27b; source_default:default@qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 47.5 — qwen-3-8-27b; scaffolded:harness-loop-truncated-tools-agent+systemid-apex-agents-v1-1-loop-truncated-tools-agent+xhigh@qwen-3-8-27b; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 48.553 — alibaba/qwen3.8-27b; first_party:xhigh@alibaba-qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.6825 — qwen3.8-27b; first_party:xhigh@qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 68.52633333333333 — qwen3.8-27b; source_default:default@qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 83.12075 — qwen3.8-27b; source_default:default@qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 73.50195833333333 — qwen3.8-27b; source_default:default@qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1363.5107335330222 — qwen3.8-27b; source_default:default@qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1242.7711821901325 — qwen3.8-27b; source_default:default@qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
Creative Writing v3: 1671.3 — Qwen/Qwen3.8-27B; source_default:default@qwen-qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z
Longform Writing: 53.3 — Qwen/Qwen3.8-27B; source_default:default@qwen-qwen3-8-27b; retrieved 2026-10-04T14:23:50.120Z

1438 ±5

75.3

—

149 (147 - 152)

—

1500 ±17

1502 ±9

75.7

61.4

21.7

Open-weight Qwen3.8 model, released 2026-08-14 per the Qwen team's release notes (Hugging Face and ModelScope); on sale here since 2026-09-04. Epoch, which carried no row for it as of 2026-09-05, lists it as Qwen 3.8 27B (dated 2026-08-14) in the 2026-09-22 read. Artificial Analysis lists several effort settings; the first-listed xhigh row is quoted rather than a lower-effort sibling. The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

Kimi K2.7 Code

Moonshot AI

China

2026-06-12

yes

30.7

11 sources, 8 families, 5 operators; 32.2% budget coverage. Sensitivity rank range 35–42 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena WebDev: 1472.9007218822194 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 37.6 — kimi-k2-7-code-high; scaffolded:harness-loop-truncated-tools-agent+high+systemid-apex-agents-v1-1-loop-truncated-tools-agent@kimi-k2-7-code-high; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.3053097345132743 — kimi-k2-7-code; scaffolded:harness-mini-swe-agent+mini-swe-agent-kimi-k2-7-code-default+systemid-deepswe-v1-1-mini-swe-agent@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 59.80766666666666 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 81.206 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 67.09870833333333 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.365 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.8787878787878788 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.5403508771929825 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tier 4: 0.12195121951219512 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.21 — kimi-k2.7-code; source_default:default@kimi-k2-7-code; retrieved 2026-10-04T14:23:50.120Z

—

68.4

—

150 (148 - 152)

—

74.0

45.7

16.1

Qwen3.6-Plus

Alibaba

China

2026-04-01

yes

28.9

12 sources, 7 families, 4 operators; 37.6% budget coverage. Sensitivity rank range 36–43 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1443.4014111988229 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1460.7352331023162 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 40.846 — alibaba/qwen3.6-plus; source_default:default@alibaba-qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 59.77408333333334 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 79.775875 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 66.66566666666667 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1407.359028507704 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
SimpleQA Verified: 0.44144144144144143 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.8838383838383839 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.3824561403508772 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.17 — qwen3.6-plus; source_default:default@qwen3-6-plus; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.12 — qwen3.6-plus_none; first_party:non-reasoning@qwen3-6-plus-none; retrieved 2026-10-04T14:23:50.120Z

1443 ±4

68.9

—

148 (145 - 149)

—

1482 ±12

1496 ±6

78.2

41.4

—

The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

MiniMax-M3

MiniMax

China

2026-06-01

yes

28.7

13 sources, 6 families, 6 operators; 43.6% budget coverage. Sensitivity rank range 38–43 across 66 recomputation scenarios (not a confidence interval).

Source observations

Arena Text Style-Controlled: 1440.0554240935571 — minimax-m3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1482.0361632685358 — minimax-m3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
Mercor APEX-Agents 1.1: 37.7 — minimax-m3; scaffolded:harness-loop-truncated-tools-agent+high+systemid-apex-agents-v1-1-loop-truncated-tools-agent@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 48.269 — minimax/MiniMax-M3; source_default:default@minimax-minimax-m3; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.6005 — MiniMax-M3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 54.42941666666667 — minimax-m3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 75.714125 — minimax-m3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 67.17195833333334 — minimax-m3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1405.9965525359048 — minimax-m3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1236.5631877536298 — minimax-m3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.9090909090909091 — MiniMax-M3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.14 — MiniMax-M3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.07 — MiniMax-M3; source_default:default@minimax-m3; retrieved 2026-10-04T14:23:50.120Z

1440 ±4

67.3

56.90

147 (143 - 150)

36.53 ±1.17

—

1473 ±11

1494 ±6

68.2

40.7

106

17.2

DeepSeek V4 Flash

DeepSeek

China

—

26.0

7 sources, 3 families, 3 operators; 25.6% budget coverage. Sensitivity rank range 41–44 across 66 recomputation scenarios (not a confidence interval). Not eligible for a rank in 6 of them.

Source observations

Arena Text Style-Controlled: 1436.0195582810732 — deepseek-v4-flash; source_default:default@deepseek-v4-flash; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1430.2571279399435 — deepseek-v4-flash-high-preview; source_default:default@deepseek-v4-flash-high-preview; retrieved 2026-10-04T14:23:50.120Z
DeepSWE v1.1: 0.5331858407079646 — deepseek-v4-flash; scaffolded:harness-mini-swe-agent+max+mini-swe-agent-deepseek-v4-flash-max+systemid-deepswe-v1-1-mini-swe-ag@deepseek-v4-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 编程综合: 53.42700000000001 — deepseek-v4-flash; source_default:default@deepseek-v4-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 75.114625 — deepseek-v4-flash; source_default:default@deepseek-v4-flash; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 66.63075 — deepseek-v4-flash; source_default:default@deepseek-v4-flash; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1408.1173107186996 — deepseek-v4-flash; source_default:default@deepseek-v4-flash; retrieved 2026-10-04T14:23:50.120Z

1436 ±4

—

146 (144 - 148)

—

1471 ±11

1484 ±6

—

The pre-0731 build. LMArena still lists this build and carries no 0731 row in its 2026-10-02 dataset publish; LiveBench dropped its row between the 2026-08-19 and 2026-08-26 snapshots, so its 65.5 is no longer quoted. Epoch lists this build as DeepSeek-V4-Flash, dated 2026-04-24, separately from its DeepSeek V4 Flash 0731 row.

—

Hy4-preview

Tencent

China

2026-08-28

yes

—

Unranked: insufficient categories. 3 sources, 3 families, 3 operators; 9.0% budget coverage.

Source observations

Arena WebDev: 1633.2225450959525 — hy4-preview; source_default:default@hy4-preview; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 55.063 — tencent/hy4-preview; source_default:default@tencent-hy4-preview; retrieved 2026-10-04T14:23:50.120Z
TapTap Maker Benchmark: 0.8095 — hy4-preview; first_party:high@hy4-preview; retrieved 2026-10-04T14:23:50.120Z

—

49.94 ±1.15

—

Tencent announced and open-sourced Hy4 preview on 2026-08-28. The Vals Index lists it as Hy4 Preview (tencent/hy4-preview); Artificial Analysis, LMArena, LiveBench and Epoch carried no row for it as of 2026-09-21, and those figures are left blank. Hy3 scores and AutoEval results are not attributed to this row.

—

DeepSeek V4 Flash Vision Exp

DeepSeek

China

2026-08-21

—

Unranked: insufficient families, operators. 3 sources, 1 families, 1 operators; 9.6% budget coverage.

Source observations

LiveBench · 编程综合: 66.65174999999999 — deepseek-v4-flash-vision-exp; source_default:default@deepseek-v4-flash-vision-exp; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 推理与数学: 86.602875 — deepseek-v4-flash-vision-exp; source_default:default@deepseek-v4-flash-vision-exp; retrieved 2026-10-04T14:23:50.120Z
LiveBench · 语言与指令: 75.65895833333333 — deepseek-v4-flash-vision-exp; source_default:default@deepseek-v4-flash-vision-exp; retrieved 2026-10-04T14:23:50.120Z

—

76.8

—

68.2

65.1

231

—

DeepSeek's experimental vision tier, announced in the vendor's change log on 2026-08-21 as model='deepseek-v4-flash-vision-exp'. LiveBench names its row with the vendor's -Exp suffix and Artificial Analysis drops it, but DeepSeek publishes only one vision tier, so both rows are this model. LMArena, Vals, SuperCLUE and Epoch carry no row for it, and the sibling flash figures are not reused. DeepSeek retired this tier on 2026-09-10 and its compatibility ID now routes to V4.1 Flash. It is absent from the public catalogue read on 2026-09-12 and is retained here as a historical evaluation row without an on-sale badge.

—

Hunyuan Hy3

Tencent

China

2026-07-06

yes

—

Unranked: insufficient families, operators, categories. 3 sources, 1 families, 1 operators; 12.4% budget coverage.

Source observations

Arena Text Style-Controlled: 1455.9834488321558 — hy3; source_default:default@hy3; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1507.4478218115694 — hy3; source_default:default@hy3; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1421.9221363428774 — hy3; source_default:default@hy3; retrieved 2026-10-04T14:23:50.120Z

1456 ±7

—

62.13

—

1509 ±21

1500 ±11

—

LongCat-2.0

Meituan

China

2026-06-30

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

54.22

—

Artificial Analysis no longer printed an output-speed value for this row on 2026-09-12, so the old figure was removed rather than retained without a live source.

—

Doubao-Seed-2.1-pro

ByteDance

China

2026-06-28

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

65.94

—

Terminal Bench 2.1 71.0

—

Doubao-Seed-2.1-turbo

ByteDance

China

2026-06-28

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

Terminal Bench 2.1 67.6

—

No independent board in this snapshot covers it; the figure shown is the vendor's own. The vendor writes the name without the 260628 snapshot suffix we serve it under.

—

Step 3.7 Flash

StepFun

China

2026-05-29

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

19*

48.16

—

193

—

Qwen3.6-Flash

Alibaba

China

2026-04-26

yes

—

Unranked: insufficient operators, categories. 5 sources, 4 families, 1 operators; 12.6% budget coverage.

Source observations

SimpleQA Verified: 0.15915915915915915 — qwen3.6-flash; source_default:default@qwen3-6-flash; retrieved 2026-10-04T14:23:50.120Z
GPQA Diamond: 0.8333333333333334 — qwen3.6-flash; source_default:default@qwen3-6-flash; retrieved 2026-10-04T14:23:50.120Z
FrontierMath v2 · Tiers 1–3: 0.17192982456140352 — qwen3.6-flash; source_default:default@qwen3-6-flash; retrieved 2026-10-04T14:23:50.120Z
Chess Puzzles: 0.2 — qwen3.6-flash; source_default:default@qwen3-6-flash; retrieved 2026-10-04T14:23:50.120Z
Mystery Game Puzzles: 0.12 — qwen3.6-flash; source_default:default@qwen3-6-flash; retrieved 2026-10-04T14:23:50.120Z

—

143 (141 - 145)

—

MiMo-V2.5

Xiaomi

China

2026-04-22

yes

—

Unranked: insufficient families, operators. 5 sources, 2 families, 2 operators; 21.4% budget coverage.

Source observations

Arena Text Style-Controlled: 1433.6058125335762 — mimo-v2.5; source_default:default@mimo-v2-5; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1437.5705946764997 — mimo-v2.5; source_default:default@mimo-v2-5; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 36.728 — xiaomi/mimo-v2.5; source_default:default@xiaomi-mimo-v2-5; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1394.3735497754396 — mimo-v2.5; source_default:default@mimo-v2-5; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1233.1834299544232 — mimo-v2.5; source_default:default@mimo-v2-5; retrieved 2026-10-04T14:23:50.120Z

1434 ±4

—

25*

—

1479 ±12

1491 ±7

—

The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

—

Step 3.5 Flash 2603

StepFun

China

2026-04-02

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

17*

—

118

—

StepFun's agent-optimised variant of step-3.5-flash, whose reasoning_effort takes only low and high. Released 2026-04-02 per StepFun's announcement on its X account ("Step 3.5 Flash 2603 just shipped!"). Artificial Analysis lists it as Step 3.5 Flash 2603 (released 2026-04-02 on its model page) and marks the row deprecated; the leaderboard hides such rows under its default Status: Current filter, so its figures are read with the Status filter set to All, and its Intelligence Index is an estimate (*). LMArena carries no 2603 row: its step-3.5-flash row is the base build and is not reused here. LiveBench, Epoch, Vals and SuperCLUE carry no row for it either.

—

MiniMax-M2.7

MiniMax

China

2026-03-18

yes

—

Unranked: insufficient families, operators, categories. 4 sources, 2 families, 2 operators; 15.4% budget coverage.

Source observations

Arena Text Style-Controlled: 1414.9801854507657 — minimax-m2.7; source_default:default@minimax-m2-7; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1396.9754832366198 — minimax-m2.7; source_default:default@minimax-m2-7; retrieved 2026-10-04T14:23:50.120Z
Vals Finance Agent: 27.887 — minimax/MiniMax-M2.7; source_default:default@minimax-minimax-m2-7; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1364.048797942928 — minimax-m2.7; source_default:default@minimax-m2-7; retrieved 2026-10-04T14:23:50.120Z

1415 ±3

—

146 (138 - 148)

—

1447 ±9

1480 ±5

—

The previously cited Vals Index V2 row is absent from the V2.1 board read on 2026-10-01, so its old V2 figure is no longer quoted.

—

GLM-5-Turbo

Zhipu AI (Z.ai)

China

2026-03-16

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

No independent board in this snapshot carries a row for it.

—

Step 3.5 Flash

StepFun

China

2026-02-02

yes

—

Unranked: insufficient sources, families, operators, categories. 2 sources, 1 families, 1 operators; 10.0% budget coverage.

Source observations

Arena Text Style-Controlled: 1393.2737669018027 — step-3.5-flash; source_default:default@step-3-5-flash; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1345.4302909079101 — step-3.5-flash; source_default:default@step-3-5-flash; retrieved 2026-10-04T14:23:50.120Z

1393 ±4

—

17*

—

1432 ±10

1449 ±6

—

126

—

StepFun's platform documentation lists step-3.5-flash as the base version and step-3.5-flash-2603 as an agent-optimised variant built from it; the 2603 build shipped on 2026-04-02 and has its own row here. LMArena added step-3.5-flash to its text board on 2026-02-10, linking the original open weights, seven weeks before 2603 shipped, so its three text-board rows are this build (60,207 votes on the overall board in its 2026-10-02 dataset publish). Artificial Analysis lists this build as Step 3.5 Flash (released 2026-02-02 on its model page) and marks the row deprecated; the leaderboard hides such rows under its default Status: Current filter, so its figures are read with the Status filter set to All, and its Intelligence Index is an estimate (*). The vendor's own SWE-bench Verified figure (74.4) that stood here while no independent board was quoted has been removed. LiveBench, Epoch, Vals and SuperCLUE carry no row for it.

—

Agnes 2.0 Flash

Agnes AI

China

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

Claw-Eval Pass^3 60.9%

—

No independent board in this snapshot covers it; the figure shown is the vendor's own. Retired from sale 2026-09-03 (CA-654): the vendor deprecated it in favour of agnes-2.5-flash; kept as a ranked row with on_chinaapi false so the roster count and maintained.at still describe the 2026-08-31 board read.

—

Agnes 2.5 Flash

Agnes AI

China

—

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

Artificial Analysis carries a different build of this family (Agnes 2.5 Pro Alpha, 40) and the vendor publishes its own figures only as a chart image, so there is no number here that can be quoted.

—

Agnes 3.0 Flash

Agnes AI

China

—

yes

—

Unranked: insufficient sources, families, operators, categories, directAnchors. 0 sources, 0 families, 0 operators; 0.0% budget coverage.

Source observations

—

36*

—

Artificial Analysis lists Agnes 3.0 Flash separately from the 2.5 family. Its Intelligence Index is currently estimated, so the trailing * is retained exactly as the board prints it. Artificial Analysis no longer printed an output-speed value for this row on 2026-09-21 (the column reads --), so the 238 tokens/s figure was removed rather than retained without a live source; the index, GDPval-AA and Terminal-Bench figures are from the 2026-09-21 board read.

—

Doubao-Seed-Evolving

ByteDance

China

—

yes

—

Unranked: insufficient sources, families, operators, categories. 1 sources, 1 families, 1 operators; 3.6% budget coverage.

Source observations

TapTap Maker Benchmark: 0.7831 — doubao-seed-evolving; first_party:high@doubao-seed-evolving; retrieved 2026-10-04T14:23:50.120Z

—

Doubao publishes Seed Evolving as a rolling model under the stable doubao-seed-evolving API ID, so capabilities can change without a dated version suffix. No independent score could be verified for this exact model on the boards cited by this snapshot as of 2026-09-12; Seed 2.1 scores are not reused.

—

GLM-5V-Turbo

Zhipu AI (Z.ai)

China

—

yes

—

Unranked: insufficient families, operators, categories. 4 sources, 1 families, 1 operators; 18.4% budget coverage.

Source observations

Arena Text Style-Controlled: 1433.4849052234497 — glm-5v-turbo; source_default:default@glm-5v-turbo; retrieved 2026-10-04T14:23:50.120Z
Arena WebDev: 1399.7850652699487 — glm-5v-turbo; source_default:default@glm-5v-turbo; retrieved 2026-10-04T14:23:50.120Z
Arena Creative Writing: 1405.741540133535 — glm-5v-turbo; source_default:default@glm-5v-turbo; retrieved 2026-10-04T14:23:50.120Z
Arena Vision Style-Controlled: 1230.0856797278223 — glm-5v-turbo; source_default:default@glm-5v-turbo; retrieved 2026-10-04T14:23:50.120Z

1433 ±6

—

1474 ±21

1488 ±11

—

Vision-language model. LMArena's text board now carries a glm-5v-turbo row (10,436 votes in its 2026-10-02 dataset publish); Epoch does not score it.

—

Qwen3.8-2.4T-A95B

Alibaba

China

—

yes

—

Unranked: insufficient sources, families, operators, categories. 1 sources, 1 families, 1 operators; 3.6% budget coverage.

Source observations

Creative Writing v3: 1842.5 — Qwen/Qwen3.8-2.4T-A95B; source_default:default@qwen-qwen3-8-2-4t-a95b; retrieved 2026-10-04T14:23:50.120Z

—

0.0

The open-weight Qwen3.8-2.4T-A95B release has its own Artificial Analysis row. Qwen identifies Qwen3.8-Max as a production version based on this model with additional features; the two catalogue models remain separate here, and Max scores from other boards are not reused. Artificial Analysis prints MLCR-AA 0.0 for this release, with no breakdown, against 19.4 and 20.0 for the two Qwen3.8-Max builds.

Image generation models

Quality coverage: 13/17 models, 4 boards from 2 evaluation operators. Public preference evidence, not comprehensive capability certification or a ChinaAPI endpoint test. Tasks and audio pools are separate; scores across columns are not interchangeable. Overlapping intervals do not establish a significant lead.

Ordered by LMArena Text-to-Image Arena; a model with no figure on that board follows the ones that have one, because absent from a board is not the same as last on it.

Image generation models, public benchmark results as of 2026-10-05
#	Model	Vendor	Origin	Released	On ChinaAPI	LMArena Text-to-Image Arena	LMArena Image Edit Arena	Artificial Analysis Text To Image	Artificial Analysis Editing	Notes
1	GPT Image 2.5 Sunburst	OpenAI	International	2026-09-08	yes	1424 ±8 — gpt-image-2.5-sunburst; text_to_image; 95% CI; samples 10884; retrieved 2026-10-05	1522 ±5 — gpt-image-2.5-sunburst; image_editing; 95% CI; samples 47289; retrieved 2026-10-05	1179 -11/11 — GPT Image 2.5 Sunburst (max); text_to_image; 95% CI; samples 5159; retrieved 2026-09-13	1167 -10/10 — GPT Image 2.5 Sunburst (max); image_editing; 95% CI; samples 5216; retrieved 2026-09-13	Sunburst and Flare are distinct builds; their scores are not interchangeable.
2	GPT Image 2.5 Flare	OpenAI	International	2026-09-08	yes	1401 ±8 — gpt-image-2.5-flare; text_to_image; 95% CI; samples 10058; retrieved 2026-10-05	1478 ±5 — gpt-image-2.5-flare; image_editing; 95% CI; samples 44792; retrieved 2026-10-05	1187 -11/11 — GPT Image 2.5 Flare (max); text_to_image; 95% CI; samples 5236; retrieved 2026-09-13	1145 -10/10 — GPT Image 2.5 Flare (max); image_editing; 95% CI; samples 5319; retrieved 2026-09-13	Flare and Sunburst are distinct builds; their scores are not interchangeable.
3	GPT Image 2	OpenAI	International	2026-04-21	yes	1383 ±4 — gpt-image-2 (medium); text_to_image; 95% CI; samples 88744; retrieved 2026-10-05	1461 ±3 — gpt-image-2 (medium); image_editing; 95% CI; samples 293819; retrieved 2026-10-05	1171 -9/9 — GPT Image 2 (high); text_to_image; 95% CI; samples 15419; retrieved 2026-09-13	1114 -9/9 — GPT Image 2 (high); image_editing; 95% CI; samples 13425; retrieved 2026-09-13	Arena quotes medium quality; Artificial Analysis quotes high quality. These are different test configurations, not interchangeable scores.
4	Gemini 3.1 Flash Image	Google	International	—	yes	1261 ±4 — gemini-3.1-flash-image (nano-banana-2) [web-search]; text_to_image; 95% CI; samples 53228; retrieved 2026-10-05	1387 ±3 — gemini-3.1-flash-image (nano-banana-2) [web-search]; image_editing; 95% CI; samples 216640; retrieved 2026-10-05	1122 -8/8 — Nano Banana 2 (Gemini 3.1 Flash Image); text_to_image; 95% CI; samples 16711; retrieved 2026-09-13	1103 -9/9 — Nano Banana 2 (Gemini 3.1 Flash Image); image_editing; 95% CI; samples 13522; retrieved 2026-09-13	Arena identifies a web-search-enabled entry; Artificial Analysis does not identify that condition in its row label. Do not assume identical configurations.
5	Doubao Seedream 5.0 Pro	ByteDance	China	—	yes	1256 ±4 — seedream-5.0-pro; text_to_image; 95% CI; samples 85690; retrieved 2026-10-05	1394 ±3 — seedream-5.0-pro; image_editing; 95% CI; samples 270020; retrieved 2026-10-05	1081 -8/8 — Seedream 5.0 Pro; text_to_image; 95% CI; samples 12814; retrieved 2026-09-13	1100 -9/9 — Seedream 5.0 Pro; image_editing; 95% CI; samples 12053; retrieved 2026-09-13	The 260628 Pro SKU is distinct from the 260128 Lite SKU. Scores quote the arena's seedream-5.0-pro rows, not seedream-5.0-lite.
6	Qwen Image 3.0 Pro	Alibaba	China	—	yes	1256 ±6 — qwen-image-3.0-pro; text_to_image; 95% CI; samples 11520; retrieved 2026-10-05	—	1084 -10/10 — Qwen-Image-3.0-Pro; text_to_image; 95% CI; samples 5896; retrieved 2026-09-13	1076 -9/9 — Qwen-Image-3.0-Pro; image_editing; 95% CI; samples 12051; retrieved 2026-09-13	No Qwen Image 3.0 Pro row on the Arena image-edit board as read from LMArena's dataset on 2026-10-05. The qwen-image-2.0-pro-2026-06-22 row is an older build and is not reused here.
7	FLUX.2 Pro	Black Forest Labs	International	2025-11-25	no	1154 ±3 — flux-2-pro; text_to_image; 95% CI; samples 205540; retrieved 2026-10-05	1245 ±3 — flux-2-pro; image_editing; 95% CI; samples 624352; retrieved 2026-10-05	1009 -8/8 — FLUX.2 [pro]; text_to_image; 95% CI; samples 9794; retrieved 2026-09-13	1006 -7/7 — FLUX.2 [pro]; image_editing; 95% CI; samples 8015; retrieved 2026-09-13
8	Doubao Seedream 4.5	ByteDance	China	2025-11-28	yes	1147 ±3 — seedream-4.5; text_to_image; 95% CI; samples 287128; retrieved 2026-10-05	1302 ±2 — seedream-4.5; image_editing; 95% CI; samples 1266531; retrieved 2026-10-05	1008 -9/9 — Seedream 4.5; text_to_image; 95% CI; samples 6308; retrieved 2026-09-13	1035 -10/10 — Seedream 4.5; image_editing; 95% CI; samples 4166; retrieved 2026-09-13
9	Doubao Seedream 5.0 Lite	ByteDance	China	2026-01-28	yes	1138 ±3 — seedream-5.0-lite; text_to_image; 95% CI; samples 123178; retrieved 2026-10-05	1294 ±3 — seedream-5.0-lite; image_editing; 95% CI; samples 446300; retrieved 2026-10-05	999 -9/9 — Seedream 5.0 Lite; text_to_image; 95% CI; samples 5027; retrieved 2026-09-13	1050 -10/10 — Seedream 5.0 Lite; image_editing; 95% CI; samples 4912; retrieved 2026-09-13	The 260128 API SKU is Seedream 5.0 Lite. Scores quote the arena's seedream-5.0-lite rows; the distinct 260628 Pro SKU has its own row. The original rank anchor is preserved.
10	Wan 2.7 Image Pro	Alibaba	China	2026-04-01	yes	1103 ±5 — wan2.7-image-pro; text_to_image; 95% CI; samples 30229; retrieved 2026-10-05	1303 ±4 — wan2.7-image-pro; image_editing; 95% CI; samples 43992; retrieved 2026-10-05	980 -9/9 — Wan 2.7 Pro; text_to_image; 95% CI; samples 5059; retrieved 2026-09-13	1039 -10/10 — Wan 2.7 Pro; image_editing; 95% CI; samples 4833; retrieved 2026-09-13
11	Wan 2.7 Image	Alibaba	China	2026-04-01	yes	1101 ±5 — wan2.7-image; text_to_image; 95% CI; samples 30621; retrieved 2026-10-05	1301 ±4 — wan2.7-image; image_editing; 95% CI; samples 44814; retrieved 2026-10-05	977 -9/9 — Wan 2.7; text_to_image; 95% CI; samples 5029; retrieved 2026-09-13	1033 -9/9 — Wan 2.7; image_editing; 95% CI; samples 6291; retrieved 2026-09-13
12	Step Image Edit 2	StepFun	China	—	no	—	—	810 -10/10 — Step Image Edit 2; text_to_image; 95% CI; samples 2593; retrieved 2026-09-13	1008 -10/10 — Step Image Edit 2; image_editing; 95% CI; samples 4235; retrieved 2026-09-13	No row on the LMArena image boards at the time of this snapshot; the arena carries an older StepFun build (step1x-edit, 998) that is a different model. Artificial Analysis evidence is now quoted separately for the identified task and exact entry.
13	Agnes Image 2.0 Flash	Agnes AI	China	—	yes	—	—	—	1017 -10/10 — Agnes Image 2.0 Flash; image_editing; 95% CI; samples 3713; retrieved 2026-09-13	No row on the LMArena image boards at the time of this snapshot. Artificial Analysis evidence is now quoted separately for the identified task and exact entry.
14	Agnes Image 2.1 Flash	Agnes AI	China	—	yes	—	—	—	—	No row on the LMArena image boards at the time of this snapshot.
15	Agnes Image 2.5 Flash	Agnes AI	China	—	yes	—	—	—	—	No Agnes Image 2.5 Flash row on either Arena image board as read from LMArena's dataset on 2026-10-05; no independent score is substituted.
16	MiniMax image-01	MiniMax	China	—	no	—	—	—	—	No row on the LMArena image boards at the time of this snapshot.
17	Step 2X Large	StepFun	China	—	no	—	—	—	—	No row on the LMArena image boards at the time of this snapshot; the arena carries an older StepFun build (step1x-edit, 998) that is a different model.

Third-party price and generation-time references

Not ChinaAPI prices or endpoint measurements. These references never affect quality coverage or ranking. Compare only matching tasks and configurations. Generation time is a trailing-three-day median of successful requests including download, not retry overhead.

Model	Task	Metric	Value	Evidence
GPT Image 2.5 Sunburst	text_to_image	price	210.7 USD / 1,000 images	GPT Image 2.5 Sunburst (max) — OpenAI; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
GPT Image 2.5 Sunburst	image_editing	price	210.7 USD / 1,000 images	GPT Image 2.5 Sunburst (max) — OpenAI; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
GPT Image 2.5 Sunburst	text_to_image	generation_time	112.476080179214 seconds	GPT Image 2.5 Sunburst (max), OpenAI — OpenAI; 1024×1024 or nearest supported; 1 image; 3-day median; includes download; Artificial Analysis; retrieved 2026-09-13
GPT Image 2.5 Flare	text_to_image	price	210.7 USD / 1,000 images	GPT Image 2.5 Flare (max) — OpenAI; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
GPT Image 2.5 Flare	image_editing	price	210.7 USD / 1,000 images	GPT Image 2.5 Flare (max) — OpenAI; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
GPT Image 2.5 Flare	text_to_image	generation_time	53.0124175548553 seconds	GPT Image 2.5 Flare (max), OpenAI — OpenAI; 1024×1024 or nearest supported; 1 image; 3-day median; includes download; Artificial Analysis; retrieved 2026-09-13
GPT Image 2	text_to_image	price	211.0 USD / 1,000 images	GPT Image 2 (high) — OpenAI; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
GPT Image 2	image_editing	price	211.0 USD / 1,000 images	GPT Image 2 (high) — OpenAI; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Gemini 3.1 Flash Image	text_to_image	price	67.0 USD / 1,000 images	Nano Banana 2 (Gemini 3.1 Flash Image) — Google; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Gemini 3.1 Flash Image	image_editing	price	67.0 USD / 1,000 images	Nano Banana 2 (Gemini 3.1 Flash Image) — Google; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Doubao Seedream 5.0 Pro	text_to_image	price	90.0 USD / 1,000 images	Seedream 5.0 Pro — ByteDance Seed; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Doubao Seedream 5.0 Pro	image_editing	price	90.0 USD / 1,000 images	Seedream 5.0 Pro — ByteDance Seed; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Qwen Image 3.0 Pro	text_to_image	price	40.0 USD / 1,000 images	Qwen-Image-3.0-Pro — Alibaba; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Qwen Image 3.0 Pro	image_editing	price	40.0 USD / 1,000 images	Qwen-Image-3.0-Pro — Alibaba; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
FLUX.2 Pro	text_to_image	price	30.0 USD / 1,000 images	FLUX.2 [pro] — Black Forest Labs; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
FLUX.2 Pro	image_editing	price	45.0 USD / 1,000 images	FLUX.2 [pro] — Black Forest Labs; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
FLUX.2 Pro	text_to_image	generation_time	11.2778015136719 seconds	FLUX.2 [pro], Black Forest Labs — Black Forest Labs; 1024×1024 or nearest supported; 1 image; 3-day median; includes download; Artificial Analysis; retrieved 2026-09-13
Doubao Seedream 4.5	text_to_image	price	40.0 USD / 1,000 images	Seedream 4.5 — ByteDance Seed; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Doubao Seedream 4.5	image_editing	price	40.0 USD / 1,000 images	Seedream 4.5 — ByteDance Seed; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Doubao Seedream 5.0 Lite	text_to_image	price	35.0 USD / 1,000 images	Seedream 5.0 Lite — ByteDance Seed; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Doubao Seedream 5.0 Lite	image_editing	price	35.0 USD / 1,000 images	Seedream 5.0 Lite — ByteDance Seed; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Wan 2.7 Image Pro	text_to_image	price	64.0 USD / 1,000 images	Wan 2.7 Pro — Alibaba; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Wan 2.7 Image	text_to_image	price	26.0 USD / 1,000 images	Wan 2.7 — Alibaba; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Wan 2.7 Image	image_editing	price	26.0 USD / 1,000 images	Wan 2.7 — Alibaba; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Step Image Edit 2	text_to_image	price	3.0 USD / 1,000 images	Step Image Edit 2 — StepFun; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Step Image Edit 2	image_editing	price	3.0 USD / 1,000 images	Step Image Edit 2 — StepFun; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13
Agnes Image 2.0 Flash	image_editing	price	3.0 USD / 1,000 images	Agnes Image 2.0 Flash — Sapiens AI; 1024×1024; creator API; default settings (AA pricing footnote); Artificial Analysis; retrieved 2026-09-13

Video generation models

Quality coverage: 15/18 models, 4 boards from 2 evaluation operators. Public preference evidence, not comprehensive capability certification or a ChinaAPI endpoint test. Tasks and audio pools are separate; scores across columns are not interchangeable. Overlapping intervals do not establish a significant lead.

Ordered by LMArena Text-to-Video Arena; a model with no figure on that board follows the ones that have one, because absent from a board is not the same as last on it.

Video generation models, public benchmark results as of 2026-10-05
#	Model	Vendor	Origin	Released	On ChinaAPI	LMArena Text-to-Video Arena	LMArena Image-to-Video Arena	Artificial Analysis Text To Video · With Audio	Artificial Analysis Image To Video · With Audio	Notes
1	Gemini Omni Flash (video)	Google	International	—	no	1513 ±9 — gemini-omni-flash; text_to_video; 95% CI; samples 26576; retrieved 2026-10-05	1465 ±6 — gemini-omni-flash; image_to_video; 95% CI; samples 87155; retrieved 2026-10-05	1238 -6/6 — Gemini Omni Flash; text_to_video_audio; 95% CI; samples 16372; retrieved 2026-09-13	1181 -7/7 — Gemini Omni Flash; image_to_video_audio; 95% CI; samples 11592; retrieved 2026-09-13
2	Doubao Seedance 2.0	ByteDance	China	2026-01-28	yes	1479 ±8 — dreamina-seedance-2.0-720p; text_to_video; 95% CI; samples 56318; retrieved 2026-10-05	1475 ±7 — dreamina-seedance-2.0-720p; image_to_video; 95% CI; samples 121427; retrieved 2026-10-05	1220 -6/6 — Dreamina Seedance 2.0 720p; text_to_video_audio; 95% CI; samples 22625; retrieved 2026-09-13	1196 -6/6 — Dreamina Seedance 2.0 720p; image_to_video_audio; 95% CI; samples 17072; retrieved 2026-09-13	The arena entry is the 720p Dreamina build of Seedance 2.0.
3	Wan 3.0 Video	Alibaba	China	2026-08-13	yes	1476 ±13 — wan3.0; text_to_video; 95% CI; samples 2598; retrieved 2026-10-05	1480 ±11 — wan3.0; image_to_video; 95% CI; samples 3442; retrieved 2026-10-05	1242 -9/9 — Wan 3.0; text_to_video_audio; 95% CI; samples 5746; retrieved 2026-09-13	1178 -8/8 — Wan 3.0; image_to_video_audio; 95% CI; samples 7876; retrieved 2026-09-13	Scores quote the arena's wan3.0 rows, linked to Alibaba's Wan3.0 release. These scores are not attributed to the separate Prime SKU.
4	Doubao Seedance 2.5	ByteDance	China	2026-06-28	yes	1474 ±9 — dreamina-seedance-2.5-720p; text_to_video; 95% CI; samples 8001; retrieved 2026-10-05	1477 ±7 — dreamina-seedance-2.5-720p; image_to_video; 95% CI; samples 15729; retrieved 2026-10-05	—	—	The arena entry is the 720p Dreamina build of Seedance 2.5.
5	MiniMax H3	MiniMax	China	2026-07-31	yes	1460 ±9 — minimax-h3; text_to_video; 95% CI; samples 11417; retrieved 2026-10-05	1495 ±5 — minimax-h3; image_to_video; 95% CI; samples 57112; retrieved 2026-10-05	1225 -7/7 — MiniMax H3; text_to_video_audio; 95% CI; samples 8854; retrieved 2026-09-13	1190 -8/8 — MiniMax H3; image_to_video_audio; 95% CI; samples 7461; retrieved 2026-09-13
6	Sora 2 Pro	OpenAI	International	2025-09-30	no	1368 ±7 — sora-2-pro; text_to_video; 95% CI; samples 54167; retrieved 2026-10-05	—	1076 -9/9 — Sora 2 Pro; text_to_video_audio; 95% CI; samples 4867; retrieved 2026-09-13	—
7	Veo 3.1	Google	International	2025-10-15	no	1364 ±14 — veo-3.1-audio; text_to_video; 95% CI; samples 13741; retrieved 2026-10-05	1398 ±10 — veo-3.1-audio; image_to_video; 95% CI; samples 25116; retrieved 2026-10-05	1090 -6/6 — Veo 3.1; text_to_video_audio; 95% CI; samples 8976; retrieved 2026-09-13	1088 -7/7 — Veo 3.1; image_to_video_audio; 95% CI; samples 8617; retrieved 2026-09-13
8	Wan 2.7 T2V	Alibaba	China	2026-04-03	yes	1337 ±7 — wan2.7-t2v; text_to_video; 95% CI; samples 27626; retrieved 2026-10-05	—	1104 -7/7 — Wan 2.7; text_to_video_audio; 95% CI; samples 5986; retrieved 2026-09-13	—
9	MiniMax Hailuo 2.3	MiniMax	China	2025-10-28	yes	1206 ±6 — hailuo-2.3; text_to_video; 95% CI; samples 89138; retrieved 2026-10-05	1262 ±5 — hailuo-2.3; image_to_video; 95% CI; samples 405786; retrieved 2026-10-05	—	—
10	MiniMax Hailuo 02	MiniMax	China	—	yes	1198 ±13 — hailuo-02-pro; text_to_video; 95% CI; samples 9399; retrieved 2026-10-05	1229 ±11 — hailuo-02-pro; image_to_video; 95% CI; samples 21743; retrieved 2026-10-05	—	—	Scores are the arena's hailuo-02-pro rows; the arena also carries standard and fast builds of the same family.
11	Kling v3	Kuaishou	China	2026-02-05	yes	—	1354 ±6 — kling-v3-pro; image_to_video; 95% CI; samples 206221; retrieved 2026-10-05	1108 -5/5 — Kling 3.0 1080p (Pro); text_to_video_audio; 95% CI; samples 20826; retrieved 2026-09-13	1076 -6/6 — Kling 3.0 1080p (Pro); image_to_video_audio; 95% CI; samples 16317; retrieved 2026-09-13	The arena entry is the pro build; the text-to-video board has no v3 row at the time of this snapshot.
12	Kling v3 Omni	Kuaishou	China	2026-02-05	yes	—	—	1090 -6/6 — Kling 3.0 Omni 1080p (Pro); text_to_video_audio; 95% CI; samples 8871; retrieved 2026-09-13	1066 -7/7 — Kling 3.0 Omni 1080p (Pro); image_to_video_audio; 95% CI; samples 7665; retrieved 2026-09-13	No row on the LMArena video boards at the time of this snapshot. Artificial Analysis evidence is now quoted separately for the identified task and exact entry.
13	Wan 2.7 I2V	Alibaba	China	2026-04-03	yes	—	1428 ±5 — wan2.7-i2v; image_to_video; 95% CI; samples 85521; retrieved 2026-10-05	—	1088 -7/7 — Wan 2.7; image_to_video_audio; 95% CI; samples 5703; retrieved 2026-09-13
14	HappyHorse 1.1 I2V	Alibaba	China	—	yes	—	—	—	1109 -7/7 — HappyHorse-1.1; image_to_video_audio; 95% CI; samples 13810; retrieved 2026-09-13	No HappyHorse 1.1 row on the LMArena video boards as read from LMArena's dataset on 2026-10-05. The arena carries the previous generation, happyhorse-1.0; that build's scores are deliberately not reused here. Artificial Analysis evidence is now quoted separately for the identified task and exact entry.
15	HappyHorse 1.1 T2V	Alibaba	China	—	yes	—	—	1146 -6/6 — HappyHorse-1.1; text_to_video_audio; 95% CI; samples 18445; retrieved 2026-09-13	—	No HappyHorse 1.1 row on the LMArena video boards as read from LMArena's dataset on 2026-10-05. The arena carries the previous generation, happyhorse-1.0; that build's scores are deliberately not reused here. Artificial Analysis evidence is now quoted separately for the identified task and exact entry.
16	Kling 3.0 Turbo	Kuaishou	China	—	yes	—	—	—	—	No row on the LMArena video boards at the time of this snapshot.
17	MiniMax Hailuo 2.3 Fast	MiniMax	China	2025-10-28	yes	—	—	—	—	No row on the LMArena video boards at the time of this snapshot.
18	Wan 3.0 Video Prime	Alibaba	China	—	yes	—	—	—	—	No separately identified Wan 3.0 Prime row on either Arena video board as read from LMArena's dataset on 2026-10-05. The wan3.0 row does not identify Prime, so its scores are not reused here.

Third-party price and generation-time references

Model	Task	Metric	Value	Evidence
Gemini Omni Flash (video)	text_to_video_audio	price	6.00 USD / video minute	Gemini Omni Flash — Google; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Gemini Omni Flash (video)	image_to_video_audio	price	6.00 USD / video minute	Gemini Omni Flash — Google; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Doubao Seedance 2.0	text_to_video_audio	price	9.07 USD / video minute	Dreamina Seedance 2.0 720p — ByteDance Seed; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Doubao Seedance 2.0	image_to_video_audio	price	9.07 USD / video minute	Dreamina Seedance 2.0 720p — ByteDance Seed; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Wan 3.0 Video	text_to_video_audio	price	12.00 USD / video minute	Wan 3.0 — Alibaba; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Wan 3.0 Video	image_to_video_audio	price	12.00 USD / video minute	Wan 3.0 — Alibaba; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
MiniMax H3	text_to_video_audio	price	7.80 USD / video minute	MiniMax H3 — MiniMax; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
MiniMax H3	image_to_video_audio	price	7.80 USD / video minute	MiniMax H3 — MiniMax; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Sora 2 Pro	text_to_video_audio	price	30.00 USD / video minute	Sora 2 Pro — OpenAI; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Veo 3.1	text_to_video_audio	price	24.00 USD / video minute	Veo 3.1 — Google; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Veo 3.1	image_to_video_audio	price	24.00 USD / video minute	Veo 3.1 — Google; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Wan 2.7 T2V	text_to_video_audio	price	9.00 USD / video minute	Wan 2.7 — Alibaba; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Kling v3	text_to_video_audio	price	20.16 USD / video minute	Kling 3.0 1080p (Pro) — KlingAI; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Kling v3	image_to_video_audio	price	20.16 USD / video minute	Kling 3.0 1080p (Pro) — KlingAI; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Kling v3	text_to_video_silent	generation_time	99.7195221020002 seconds	Kling 3.0 1080p (Pro), fal.ai — fal.ai; 720p / 5s panel target; nearest supported settings (Kling Pro: 1080p); no audio; 3-day median; includes download; Artificial Analysis; retrieved 2026-09-13
Kling v3 Omni	text_to_video_audio	price	16.80 USD / video minute	Kling 3.0 Omni 1080p (Pro) — KlingAI; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Kling v3 Omni	image_to_video_audio	price	16.80 USD / video minute	Kling 3.0 Omni 1080p (Pro) — KlingAI; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
Wan 2.7 I2V	image_to_video_audio	price	9.00 USD / video minute	Wan 2.7 — Alibaba; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
HappyHorse 1.1 I2V	image_to_video_audio	price	9.90 USD / video minute	HappyHorse-1.1 — Alibaba-ATH; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
HappyHorse 1.1 T2V	text_to_video_audio	price	9.90 USD / video minute	HappyHorse-1.1 — Alibaba-ATH; 1080p; creator API; default settings (AA pricing footnote, not the quality row resolution); Artificial Analysis; retrieved 2026-09-13
HappyHorse 1.1 T2V	text_to_video_silent	generation_time	107.870128135 seconds	HappyHorse-1.1, Alibaba Cloud (T2V) — Alibaba Cloud (T2V); 720p / 5s panel target; nearest supported settings (Kling Pro: 1080p); no audio; 3-day median; includes download; Artificial Analysis; retrieved 2026-09-13

Real usage on ChinaAPI

Which models the calls ChinaAPI actually served went to, as positions and shares of this platform's own traffic. It is a measurement of one gateway, not of the Chinese AI market, and it is not comparable with the benchmark results above. No call count, token count or revenue figure is published here or anywhere else on this site: every number below is a proportion. Recomputed hourly. A model with too little traffic in a window to rank is left unnamed and counted under “other”, as is anything outside the published catalogue. The same board as JSON is at /api/model-rankings/usage.

Share of calls served, the last 7 days

Model share of calls ChinaAPI served over the last 7 days
#	Model	Vendor	Share of calls	Trend vs the previous 7 days
1	deepseek-flash	DeepSeek	8.3%	up
2	agnes-3.0-flash	Agnes AI	7.1%	up
3	claude-opus-5-5	Anthropic	4.9%	down
4	glm-5.3-flash	Zhipu AI	4.4%	up
5	agnes-image-2.5-flash	Agnes AI	3.7%	up
6	gpt-6-sol	OpenAI	3.6%	down
7	gpt-6.1-sol	OpenAI	2.9%	flat
8	claude-haiku-5-5	Anthropic	2.1%	up
9	gpt-5.6-sol	OpenAI	2.0%	up
10	gemini-3.8-flash	Google	1.7%	down
11	claude-opus-4-8	Anthropic	1.6%	down
12	glm-5.3	Zhipu AI	1.4%	flat
13	hy3	Tencent	1.4%	down
14	deepseek-v4-pro	DeepSeek	1.4%	flat
15	claude-opus-4-6	Anthropic	1.1%	flat
16	gpt-6-astra	OpenAI	1.1%	flat
17	claude-fable-5-1	Anthropic	1.1%	flat
18	claude-opus-4-7	Anthropic	1.1%	flat
19	claude-sonnet-4-6	Anthropic	1.1%	flat
20	claude-fable-5	Anthropic	1.1%	flat
	All other models		47.1%

How those calls arrived: 18.9% over agent protocols (Claude Messages and the OpenAI Responses API), 74.0% over chat protocols (OpenAI chat completions and Gemini generateContent), 7.2% over everything else (audio, images, embeddings, rerank and asynchronous tasks).

Share of calls served, the last 30 days

Model share of calls ChinaAPI served over the last 30 days
#	Model	Vendor	Share of calls	Trend vs the previous 30 days
1	hy3	Tencent	16.4%	up
2	deepseek-flash	DeepSeek	5.2%	up
3	agnes-3.0-flash	Agnes AI	3.8%	up
4	claude-opus-5-5	Anthropic	3.2%	up
5	gpt-6-sol	OpenAI	2.9%	up
6	glm-5.3-flash	Zhipu AI	2.8%	flat
7	claude-opus-4-8	Anthropic	2.1%	up
8	claude-sonnet-5	Anthropic	1.8%	flat
9	gemini-3.8-flash	Google	1.7%	up
10	gpt-5.6-sol	OpenAI	1.7%	flat
11	agnes-2.5-flash	Agnes AI	1.5%	down
12	deepseek-v4-pro	DeepSeek	1.4%	down
13	gpt-6.1-sol	OpenAI	1.3%	up
14	claude-opus-5	Anthropic	1.2%	flat
15	claude-opus-4-6	Anthropic	1.2%	flat
16	claude-sonnet-4-6	Anthropic	1.1%	flat
17	claude-opus-4-7	Anthropic	1.1%	flat
18	claude-fable-5	Anthropic	1.1%	flat
19	gpt-6-astra	OpenAI	1.0%	up
20	glm-5.3	Zhipu AI	0.9%	flat
	All other models		46.6%

How those calls arrived: 14.3% over agent protocols (Claude Messages and the OpenAI Responses API), 82.8% over chat protocols (OpenAI chat completions and Gemini generateContent), 3.0% over everything else (audio, images, embeddings, rerank and asynchronous tasks).