Chinese AI Model Rankings

Where the Chinese AI models ChinaAPI serves stand next to the international frontier, quoted from public leaderboards. Snapshot of 2026-08-10. Every figure is reproduced from the board named in its column and links back to it. ChinaAPI runs no evaluation of its own and computes no composite score from these numbers. See how this list is built, the price of every model marked available, and the same data as JSON at /api/model-rankings.

Sources

Text and multimodal models

Ordered by Epoch Capabilities Index (ECI); 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.

Text and multimodal models, public benchmark results as of 2026-08-10
#ModelVendorOriginOn ChinaAPILMArena Text ArenaLiveBenchArtificial Analysis Intelligence IndexSuperCLUE 智能指数Epoch Capabilities Index (ECI)Vals IndexVendor's own evaluation cardNotes
1Claude Fable 5AnthropicInternationalno1507 ±683.062162 (159 - 165)75.14
2GPT-5.6 SolOpenAIInternationalno1482 ±681.061162 (159 - 165)73.12
3Claude Opus 5AnthropicInternationalno1493 ±680.163161 (159 - 164)74.82
4GPT-5.5OpenAIInternationalno1482 ±480.2159 (157 - 162)67.95
5GPT-5.6 TerraOpenAIInternationalno1467 ±677.957159 (156 - 162)65.14
6Kimi K3Moonshot AIChinayes1485 ±779.26070.68157 (155 - 160)74.70
7Qwen3.8-MaxAlibabaChinayes1497 ±978.55871.48156 (154 - 158)65.47
8Gemini 3.1 ProGoogleInternationalno1487 ±377.048155 (153 - 157)53.77
9Gemini 3.6 FlashGoogleInternationalyes1485 ±673.652154 (152 - 156)62.43
10Grok 4.5xAIInternationalno1468 ±675.856154 (152 - 156)65.30LMArena labels this vendor SpaceXAI.
11Qwen3.7-MaxAlibabaChinayes1475 ±1073.1154 (152 - 156)57.49LMArena lists the preview build (qwen3.7-max-preview); the released model has no separate row.
12DeepSeek V4 FlashDeepSeekChinayes74.25265.60153 (150 - 154)63.95The boards carry the 0731 build; LiveBench also lists an earlier DeepSeek V4 Flash row at 65.5.
13GLM-5.2Zhipu AI (Z.ai)Chinayes1471 ±573.25363.27152 (151 - 154)65.02
14GLM-5.1Zhipu AI (Z.ai)Chinayes1468 ±4151 (149 - 152)52.45
15Kimi K2.6Moonshot AIChinayes1461 ±570.5151 (149 - 153)55.17
16Kimi K2.7 CodeMoonshot AIChinayes68.443151 (149 - 152)
17DeepSeek V4 ProDeepSeekChinayes1457 ±471.64564.40149 (147 - 151)55.62The Vals Index labels this row "DeepSeek V4"; it lists the Flash build separately.
18MiniMax-M3MiniMaxChinayes67.34556.90149 (145 - 151)58.94
19Qwen3.6-PlusAlibabaChinayes68.9149 (147 - 151)48.89
20GLM-5Zhipu AI (Z.ai)Chinayes1457 ±4146 (144 - 148)
21MiniMax-M2.7MiniMaxChinayes146 (138 - 148)41.56
22Qwen3.6-FlashAlibabaChinayes144 (141 - 146)
23Agnes 2.0 FlashAgnes AIChinayesClaw-Eval Pass^3 60.9%No independent board in this snapshot covers it; the figure shown is the vendor's own.
24Agnes 2.5 FlashAgnes AIChinayesArtificial 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.
25Doubao-Seed-2.1-proByteDanceChinayes65.94Terminal Bench 2.1 71.0
26Doubao-Seed-2.1-turboByteDanceChinayesTerminal Bench 2.1 67.6No 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.
27GLM-5-TurboZhipu AI (Z.ai)ChinayesNo independent board in this snapshot carries a row for it.
28GLM-5V-TurboZhipu AI (Z.ai)ChinayesVision-language model; no independent board in this snapshot carries a row for it.
29Hunyuan Hy3TencentChinayes1453 ±104262.13
30LongCat-2.0MeituanChinayes3454.22
31MiMo-V2.5XiaomiChinayes3851.57
32MiMo-V2.5-ProXiaomiChinayes1467 ±44353.0150.74
33Muse Spark 1.2MetaInternationalno1498 ±1378.05771.88
34Qwen3.7-PlusAlibabaChinayes1458 ±53952.33
35Step 3.5 FlashStepFunChinayesSWE-bench Verified 74.4No independent board in this snapshot covers it; the figure shown is the vendor's own.
36Step 3.7 FlashStepFunChinayes3148.16

Image generation models

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 2026-08-10
#ModelVendorOriginOn ChinaAPILMArena Text-to-Image ArenaLMArena Image Edit ArenaNotes
1GPT Image 2OpenAIInternationalno1380 ±51463 ±4
2Gemini 3.1 Flash ImageGoogleInternationalno1263 ±51385 ±4
3Doubao Seedream 5.0ByteDanceChinayes1257 ±51393 ±4Scores are the arena's seedream-5.0-pro rows; the arena also carries a seedream-5.0-lite build (1136 / 1293).
4FLUX.2 ProBlack Forest LabsInternationalno1155 ±31244 ±3
5Doubao Seedream 4.5ByteDanceChinayes1146 ±31301 ±2
6Wan 2.7 Image ProAlibabaChinayes1103 ±51302 ±4
7Wan 2.7 ImageAlibabaChinayes1100 ±51301 ±4
8Agnes Image 2.0 FlashAgnes AIChinayesNo row on the LMArena image boards at the time of this snapshot.
9Agnes Image 2.1 FlashAgnes AIChinayesNo row on the LMArena image boards at the time of this snapshot.
10MiniMax image-01MiniMaxChinayesNo row on the LMArena image boards at the time of this snapshot.
11Step 2X LargeStepFunChinayesNo 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.
12Step Image Edit 2StepFunChinayesNo 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.

Video generation models

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 2026-08-10
#ModelVendorOriginOn ChinaAPILMArena Text-to-Video ArenaLMArena Image-to-Video ArenaNotes
1Gemini Omni Flash (video)GoogleInternationalno1513 ±121462 ±7
2Doubao Seedance 2.0ByteDanceChinayes1479 ±111478 ±10The arena entry is the 720p Dreamina build of Seedance 2.0.
3MiniMax H3MiniMaxChinayes1455 ±191476 ±19
4Sora 2 ProOpenAIInternationalno1365 ±8
5Veo 3.1GoogleInternationalno1364 ±141397 ±11
6Wan 2.7 T2VAlibabaChinayes1347 ±9
7MiniMax Hailuo 2.3MiniMaxChinayes1202 ±71260 ±6
8MiniMax Hailuo 02MiniMaxChinayes1198 ±131227 ±11Scores are the arena's hailuo-02-pro rows; the arena also carries standard and fast builds of the same family.
9Doubao Seedance 2.5ByteDanceChinayesNo row on the LMArena video boards at the time of this snapshot.
10HappyHorse 1.1 I2VAlibabaChinayesThe LMArena video boards carry the previous generation (happyhorse-1.0, 1428 text-to-video / 1442 image-to-video); that score is not this model's and is deliberately not reused here.
11HappyHorse 1.1 T2VAlibabaChinayesThe LMArena video boards carry the previous generation (happyhorse-1.0, 1428 text-to-video / 1442 image-to-video); that score is not this model's and is deliberately not reused here.
12Kling 3.0 TurboKuaishouChinayesNo row on the LMArena video boards at the time of this snapshot.
13Kling v3KuaishouChinayes1359 ±7The arena entry is the pro build; the text-to-video board has no v3 row at the time of this snapshot.
14Kling v3 OmniKuaishouChinayesNo row on the LMArena video boards at the time of this snapshot.
15MiniMax Hailuo 2.3 FastMiniMaxChinayesNo row on the LMArena video boards at the time of this snapshot.
16Wan 2.7 I2VAlibabaChinayes1427 ±7

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
#ModelVendorShare of callsTrend vs the previous 7 days
1deepseek-v4-flashDeepSeek32.3%flat
2deepseek-v4-proDeepSeek25.1%up
3mimo-v2.5Xiaomi23.3%up
4glm-5.2智谱2.1%down
5hy3Tencent1.9%up
6kimi-k3Moonshot1.3%down
7qwen3.8-max阿里巴巴1.3%up
8glm-5-turbo智谱1.2%up
9glm-5智谱1.0%flat
10kimi-k2.7-codeMoonshot0.7%flat
11glm-5.1智谱0.6%flat
12kimi-k2.6Moonshot0.4%flat
13MiniMax-M3MiniMax0.3%flat
All other models8.3%

How those calls arrived: 0.2% over agent protocols (Claude Messages and the OpenAI Responses API), 8.1% over chat protocols (OpenAI chat completions and Gemini generateContent), 91.7% 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
#ModelVendorShare of callsTrend vs the previous 30 days
1deepseek-v4-flashDeepSeek27.3%up
2deepseek-v4-proDeepSeek22.5%up
3mimo-v2.5Xiaomi18.4%up
4glm-5.2智谱7.5%up
5kimi-k3Moonshot5.9%up
6mimo-v2.5-proXiaomi2.3%up
7hy3Tencent1.4%down
8MiniMax-M2.7MiniMax1.2%up
9qwen3.6-flash阿里巴巴0.9%up
10glm-5智谱0.9%up
11glm-5.1智谱0.7%up
12glm-5-turbo智谱0.5%flat
13qwen3.7-max阿里巴巴0.5%down
14kimi-k2.7-codeMoonshot0.4%flat
15qwen3.8-max阿里巴巴0.4%flat
16kimi-k2.6Moonshot0.4%flat
17MiniMax-M3MiniMax0.4%flat
18stepaudio-2.5-asrStepFun0.2%flat
19qwen3.7-plus阿里巴巴0.2%flat
20mimo-v2.5-asrXiaomi0.2%flat
All other models7.8%

How those calls arrived: 0.1% over agent protocols (Claude Messages and the OpenAI Responses API), 2.6% over chat protocols (OpenAI chat completions and Gemini generateContent), 97.3% over everything else (audio, images, embeddings, rerank and asynchronous tasks).