🔮 MiMo Chat Completions API

📌 概述 & 基本信息

MiMo Chat Completions API 是小米大语言模型提供的 OpenAI 兼容对话补全接口，支持 RESTful HTTP 协议，所有请求与响应均使用 JSON 格式。流式响应采用 SSE（Server-Sent Events）。

核心能力一览

⚡ 快速开始

获取 API Key 后即可开始调用。支持 OpenAI SDK 兼容格式，也可直接使用 HTTP 请求。

bashcurl --location --request POST 'https://api.xiaomimimo.com/v1/chat/completions' \
--header "api-key: $MIMO_API_KEY" \
--header "Content-Type: application/json" \
--data-raw '{
    "model": "mimo-v2.5-pro",
    "messages": [
        {
            "role": "system",
            "content": "You are MiMo, an AI assistant developed by Xiaomi."
        },
        {
            "role": "user",
            "content": "你好，请介绍一下自己"
        }
    ],
    "max_completion_tokens": 1024,
    "temperature": 1.0,
    "top_p": 0.95,
    "stream": false,
    "thinking": {
        "type": "disabled"
    }
}'

pythonfrom openai import OpenAI

client = OpenAI(
    api_key="$MIMO_API_KEY",
    base_url="https://api.xiaomimimo.com/v1"
)

response = client.chat.completions.create(
    model="mimo-v2.5-pro",
    messages=[
        {"role": "system", "content": "You are MiMo, an AI assistant developed by Xiaomi."},
        {"role": "user", "content": "你好，请介绍一下自己"}
    ],
    max_completion_tokens=1024,
    temperature=1.0,
    thinking={"type": "disabled"}
)

print(response.choices[0].message.content)

💰 可用模型

🔐 认证方式

接口支持以下两种认证方式，选择其中一种添加到请求头中：

方式一：api-key 字段认证

httpapi-key: $MIMO_API_KEY
Content-Type: application/json

方式二：Authorization Bearer 认证

httpAuthorization: Bearer $MIMO_API_KEY
Content-Type: application/json

📥 请求参数详解

核心参数

生成控制参数

深度思考参数

工具调用参数

语音合成参数 (audio)

📤 响应对象（非流式输出）

当 stream: false 时，API 返回完整的 chat.completion 对象。

顶层字段

choices[] 字段

usage 用量信息

响应示例

json{
    "id": "8b51f9e0515949cb8207fbd35ea6ea5c",
    "object": "chat.completion",
    "created": 1776848906,
    "model": "mimo-v2.5-pro",
    "choices": [
        {
            "finish_reason": "stop",
            "index": 0,
            "message": {
                "content": "Hello! I'm MiMo, Xiaomi's AI assistant created by the Xiaomi LLM-Core team...",
                "role": "assistant",
                "tool_calls": null
            }
        }
    ],
    "usage": {
        "completion_tokens": 72,
        "prompt_tokens": 57,
        "total_tokens": 129,
        "completion_tokens_details": {
            "reasoning_tokens": 0
        },
        "prompt_tokens_details": null
    }
}

📤 响应 Chunk 对象（流式输出）

当 stream: true 时，API 通过 SSE 以 chat.completion.chunk 格式增量返回数据。

Chunk 特有字段（vs 非流式差异）

其余字段（id, created, model, usage, annotations 等）与非流式相同，通常仅在最后一个 chunk 中返回 usage。

🧠 深度思考模式

MiMo V2.5 Pro 和标准版支持思维链（Chain-of-Thought）推理，让模型在回答前进行深度推理。适用于数学、逻辑、编程、复杂分析等场景。

核心参数

行为说明

• 启用后，响应中 reasoning_content 字段包含模型的推理过程，content 为最终回答
• 思考模式下 temperature 强制为 1.0，top_p 强制为 0.95，自定义无效
• 多轮工具调用中，模型同时返回 tool_calls 和 reasoning_content
• 建议：后续每次请求的 messages 中保留所有历史 reasoning_content，以获得最佳表现

🔧 工具调用 (Function Calling)

MiMo 支持 Function Calling，允许模型调用外部函数获取信息或执行操作。同时内置 Web Search 联网搜索工具。

核心要点

Function Tool 结构

json{
  "type": "function",
  "function": {
    "name": "get_weather",
    "description": "获取指定城市的当前天气信息",
    "parameters": {
      "type": "object",
      "properties": {
        "location": { "type": "string", "description": "城市名称" }
      },
      "required": ["location"]
    },
    "strict": false
  }
}

🌐 联网搜索

MiMo 内置 Web Search 工具，模型可自动联网检索最新信息并在回答中引用来源。

启用方式

在 tools 数组中添加 web search 类型的工具：

json{
  "tools": [
    { "type": "web_search" }
  ]
}

返回的引用注释 (annotations)

🎙️ 语音合成 (TTS)

MiMo 提供三种 TTS 能力，通过不同的模型和 audio 参数组合实现。

三种模式对比

预置音色列表 (mimo-v2.5-tts)

音频格式

💻 代码示例

基础调用（非流式）

pythonfrom openai import OpenAI

client = OpenAI(api_key="$MIMO_API_KEY", base_url="https://api.xiaomimimo.com/v1")

response = client.chat.completions.create(
    model="mimo-v2.5-pro",
    messages=[
        {"role": "system", "content": "You are MiMo, an AI assistant developed by Xiaomi."},
        {"role": "user", "content": "你好，请介绍一下自己"}
    ],
    max_completion_tokens=1024,
    temperature=1.0,
    thinking={"type": "disabled"}
)

print(response.choices[0].message.content)
print(f"输入: {response.usage.prompt_tokens}, 输出: {response.usage.completion_tokens}")

bashcurl --location --request POST 'https://api.xiaomimimo.com/v1/chat/completions' \
--header "api-key: $MIMO_API_KEY" \
--header "Content-Type: application/json" \
--data-raw '{"model":"mimo-v2.5-pro","messages":[{"role":"system","content":"You are MiMo."},{"role":"user","content":"你好"}],"max_completion_tokens":1024,"thinking":{"type":"disabled"}}'

流式响应

pythonfrom openai import OpenAI

client = OpenAI(api_key="$MIMO_API_KEY", base_url="https://api.xiaomimimo.com/v1")

response = client.chat.completions.create(
    model="mimo-v2.5-pro",
    messages=[
        {"role": "system", "content": "你是一个专业的编程助手。"},
        {"role": "user", "content": "解释 JavaScript 的事件循环机制"}
    ],
    stream=True
)

full_reply = ""
for chunk in response:
    content = chunk.choices[0].delta.content
    if content:
        full_reply += content
        print(content, end="", flush=True)

# 继续对话
messages.append({"role": "assistant", "content": full_reply})
messages.append({"role": "user", "content": "能给一个 async/await 的代码示例吗？"})

javascriptconst response = await fetch('https://api.xiaomimimo.com/v1/chat/completions', {
  method: 'POST',
  headers: {
    'api-key': MIMO_API_KEY,
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({
    model: 'mimo-v2.5-pro',
    messages: [
      { role: 'system', content: '你是一个专业的编程助手。' },
      { role: 'user', content: '解释 JavaScript 的事件循环机制' }
    ],
    stream: true
  })
});

const reader = response.body.getReader();
const decoder = new TextDecoder();
let buffer = '';

while (true) {
  const { done, value } = await reader.read();
  if (done) break;
  buffer += decoder.decode(value, { stream: true });
  const lines = buffer.split('\n');
  buffer = lines.pop();
  for (const line of lines) {
    if (!line.startsWith('data: ') || line === 'data: [DONE]') continue;
    const chunk = JSON.parse(line.slice(6));
    const content = chunk.choices[0]?.delta?.content;
    if (content) process.stdout.write(content);
  }
}

深度思考模式

pythonfrom openai import OpenAI

client = OpenAI(api_key="$MIMO_API_KEY", base_url="https://api.xiaomimimo.com/v1")

response = client.chat.completions.create(
    model="mimo-v2.5-pro",
    messages=[{"role": "user", "content": "9.11 和 9.8 哪个更大？"}],
    extra_body={"thinking": {"type": "enabled"}}
)

msg = response.choices[0].message
print(f"[思考过程]\n{msg.reasoning_content}")
print(f"\n[最终回答]\n{msg.content}")

函数调用 (Function Calling)

pythonfrom openai import OpenAI

client = OpenAI(api_key="$MIMO_API_KEY", base_url="https://api.xiaomimimo.com/v1")

response = client.chat.completions.create(
    model="mimo-v2.5-pro",
    messages=[{"role": "user", "content": "杭州今天天气怎么样？"}],
    tools=[{
        "type": "function",
        "function": {
            "name": "get_weather",
            "description": "获取指定城市的当前天气信息",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {"type": "string", "description": "城市名称"}
                },
                "required": ["location"]
            }
        }
    }]
)

tool_calls = response.choices[0].message.tool_calls
print("工具调用:", tool_calls)

联网搜索

pythonfrom openai import OpenAI

client = OpenAI(api_key="$MIMO_API_KEY", base_url="https://api.xiaomimimo.com/v1")

response = client.chat.completions.create(
    model="mimo-v2.5-pro",
    messages=[{"role": "user", "content": "今天有什么科技新闻？"}],
    tools=[{"type": "web_search"}]
)

msg = response.choices[0].message
print("回答:", msg.content)
if msg.annotations:
    print("\n引用来源:")
    for ann in msg.annotations:
        print(f"  - [{ann.title}]({ann.url})")

语音合成 (TTS)

pythonfrom openai import OpenAI
import base64

client = OpenAI(api_key="$MIMO_API_KEY", base_url="https://api.xiaomimimo.com/v1")

response = client.chat.completions.create(
    model="mimo-v2.5-tts",
    messages=[
        {"role": "user", "content": "用甜美的声音为大家念一首诗"},
        {"role": "assistant", "content": "床前明月光，疑是地上霜。举头望明月，低头思故乡。"}
    ],
    audio={
        "format": "wav",
        "voice": "茉莉"
    }
)

audio_data = response.choices[0].message.audio.data
audio_bytes = base64.b64decode(audio_data)
with output("output.wav", "wb") as f:
    f.write(audio_bytes)
print("音频已保存到 output.wav")

⚠️ 错误码说明