Bring ChinaAPI capabilities into your agent

Choose an agent

List reviewed 2026-09-07

• Hermes Agent /v1/chat/completions
• Claude Code /v1/messages
• Cline /v1/chat/completions
• Kilo Code /v1/chat/completions
• pi /v1/chat/completions
• Codex /v1/responses
• Oh-My-Pi /v1/chat/completions
• OpenClaw /v1/messages
• DeepSeek Harness /v1/chat/completions
• OpenHands /v1/chat/completions
• Cursor /v1/chat/completions
• Zed Editor /v1/chat/completions
• Strix /v1/chat/completions
• goose /v1/chat/completions
• Qwen Code /v1/chat/completions
• GitHub Copilot /v1/chat/completions
• opencode /v1/chat/completions
• Open Interpreter /v1/chat/completions
• Crush /v1/chat/completions

Choose a capability

• Text & Multimodal
• Image
• Video
• Audio & Speech
• Decision

Recommended models

See all 95 models

Preferred for Claude Code: The models our own configuration snippet picks for this tool

The rest follow ChinaAPI Consensus v15 (snapshot 2026-10-05); models that board does not carry keep their existing order Model Capability Rankings

• kimi-k3 Preferred
  Moonshot Kimi K3 — 2.8-trillion-parameter flagship for long-horizon coding, end-to-end knowledge work, and deep reasoning, with a 1M-token context window and up to 1,048,576 completion tokens (131,072 by default). Accepts text, base64-encoded images, and base64-encoded video and returns text. Thinking is always on, with reasoning effort at max by default; temperature is fixed at 1.0. Supports custom tool calling and dynamic tool loading.
• kimi-k2.7-code Preferred
  Moonshot Kimi K2.7 Code — Kimi's dedicated coding model, which the vendor measures as improving instruction compliance and long-horizon coding over K2.6 while cutting overthinking by about 30% on average. 256K context window, 32,768 default output tokens. Accepts text, images, and video as base64 content and returns text. Thinking is mandatory and returns an error if disabled; tool_choice is limited to auto or none and reasoning_content must be carried across multi-step tool calls.
• glm-5-turbo Preferred
  Zhipu GLM-5-Turbo — the throughput-oriented GLM-5 variant, optimised for agent workflows: more stable tool and skill invocation across multi-step tasks, better handling of complex long-chain instructions, and scheduled or long-running task execution. 200K context window, up to 128K output tokens, text in and text out. Supports thinking modes, function calling, structured JSON output, streaming, context caching, and MCP tool integration.
• claude-opus-5-5
  Anthropic Claude Opus 5.5 — built for long-running agentic coding and knowledge work, at a lower price than Claude Opus 5. It accepts text and image input, has a 1,000,000-token context window, and supports up to 128,000 output tokens. Adaptive thinking is always on; the effort setting controls how deeply it reasons, with medium as the default.
• claude-fable-5-1
  Anthropic Claude Fable 5.1 — Anthropic’s latest generally available flagship for ambitious, long-running knowledge work and coding. It is built for multi-day agentic tasks, complex software engineering, deep research, document and vision reasoning, and proactive self-verification. Cache reads are priced at 2.5% of input — the lowest of its generation.
• gpt-6-astra
  OpenAI GPT-6 Astra — OpenAI's most capable model, built for the hardest end-to-end work: complex reasoning, coding, computer use, research, and document creation. It supports text and image input, a 1,050,000-token context window, up to 128,000 output tokens, and reasoning effort settings up to xhigh and max.
• claude-fable-5
  Anthropic Claude Fable 5 — Anthropic’s most capable generally available model for ambitious, long-running knowledge work and coding. It is built for multi-day agentic tasks, complex software engineering, deep research, document and vision reasoning, and proactive self-verification.
• gemini-3.7-flash
  Google Gemini 3.7 Flash — the newest Flash-tier multimodal reasoning model, priced identically to 3.6 Flash while the vendor promotion runs. Accepts text, images, video, audio and PDFs; supports thinking, function calling, code execution, search grounding and structured output. Pricing note: Google's list price for this model is promotional through December 31, 2026 — $0.75 input / $3.75 output / $0.075 cached input per 1M tokens. From January 1, 2027 the vendor's list price returns to $1.50 / $7.50 / $0.15, and our price follows it at the same parity.

Claude Code × Text & Multimodal — /v1/messages

# 1) Point Claude Code at ChinaAPI
export ANTHROPIC_BASE_URL=https://api.chinaapi.ai
export ANTHROPIC_AUTH_TOKEN=$CHINAAPI_KEY
export ANTHROPIC_DEFAULT_OPUS_MODEL=kimi-k3
export ANTHROPIC_DEFAULT_SONNET_MODEL=kimi-k2.7-code
export ANTHROPIC_DEFAULT_HAIKU_MODEL=glm-5-turbo

claude

# 2) Call Text & Multimodal — POST /v1/messages
curl -X POST https://api.chinaapi.ai/v1/messages \
  -H "x-api-key: $CHINAAPI_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{"model":"kimi-k3","max_tokens":256,"messages":[{"role":"user","content":"Hello"}]}'