OpenAI-Compatible Format — Unified LLM API Documentation

Overview#

API Endpoint#

https://api.tokenhot.ai/v1/chat/completions

Usage#

Basic Information#

Supported Models#

1. GPT (OpenAI)#

OpenAI's GPT series — the world's leading large language model family. The latest GPT-5.x series continues to push boundaries in reasoning, code generation, and multimodal capabilities, with full support for Function Calling and streaming output.

2. Claude (Anthropic)#

Anthropic's Claude series, renowned for safety, long-context understanding, and precise instruction following. The latest Claude 4.x series excels in complex reasoning, code generation, and multilingual tasks.

3. Gemini (Google)#

Google's Gemini series with native multimodal architecture, supporting text, image, audio, video, and more. The latest Gemini 3.x series further enhances reasoning and tool-use capabilities.

4. Qwen — Alibaba Cloud#

Alibaba Cloud's Qwen series, covering flagship, balanced, and high-speed tiers with full support for Function Calling and streaming output. The latest Qwen3.6 series further improves reasoning and multimodal capabilities with million-token context support.

5. DeepSeek#

DeepSeek series models. The latest V4 series achieves major breakthroughs in inference efficiency and generation quality. V3.2 introduced the Sparse Attention mechanism (DSA), significantly reducing inference costs and improving long-context processing.

⚠️ Notes for deepseek-reasoner:

Does not support temperature, top_p, presence_penalty, or frequency_penalty parameters

Does not support Function Calling

Responses will include an additional reasoning_content field

For multi-turn conversations, remove reasoning_content from message history

6. xAI (Grok)#

xAI's Grok series. The latest Grok 4.x series features multi-agent collaborative architecture with support for ultra-long context and deep reasoning.

7. Zhipu AI (GLM)#

Zhipu AI's GLM series. The latest GLM-5.1 further upgrades reasoning depth and instruction following. GLM-5 features Dynamic Sparse Attention (DSA) with excellent performance in conversation, coding, and agent tasks.

8. MiniMax#

MiniMax series models. The latest M2.7 series focuses on agent workflows and advanced reasoning, supporting both OpenAI and Anthropic protocols.

9. Moonshot / Kimi#

Moonshot AI's Moonshot and Kimi series, renowned for ultra-long context processing and agent capabilities. The latest Kimi K2.6 further improves reasoning and multimodal capabilities, while K2.5 supports native multimodal and chain-of-thought reasoning.

Request Examples#

Python Example#

Tip: Replace sk-********************************** with your actual API Key before use. Get your key at: https://tokenhot.ai/api-key

cURL Example#

OpenAI SDK Example (Python)#

Node.js Example#

Response Example#

Successful Response Structure#

{
  "id": "chatcmpl-abc123def456",
  "object": "chat.completion",
  "created": 1711712000,
  "model": "gpt-5.5",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Quantum computing is a new computing paradigm based on the principles of quantum mechanics. Unlike classical computers that use bits (0 or 1), quantum computers use quantum bits (qubits), which can exist in a superposition of both 0 and 1 simultaneously..."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 28,
    "completion_tokens": 156,
    "total_tokens": 184
  }
}

Response Field Descriptions#

Request Parameters#

Message Roles in the messages Array#

Error Codes and Error Responses#

Error Code Overview#

400 — Business Error (BusinessError)#

{
  "code": "video_url_required",
  "message": "model doubao-seedance-2.0-V2V requires video_url content",
  "data": null
}

401 / 503 — Gateway Error (GatewayError)#

{
  "error": {
    "code": "",
    "message": "Invalid token (request id: 20260327...)",
    "type": "new_api_error"
  }
}

{
  "error": {
    "code": "model_not_found",
    "message": "No available channel for the current group (request id: 20260330...)",
    "type": "new_api_error"
  }
}

Note: Different models may have slight variations in parameter support (e.g., deepseek-reasoner does not support temperature). Please refer to each model's detailed description for specifics. If you have any questions, please contact TokenHot support.