> ## Documentation Index
> Fetch the complete documentation index at: https://platform.minimaxi.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Chat Completions API
> 使用 OpenAI API 兼容 Chat Completions 格式调用 MiniMax 模型。
✨ **全新模型 `MiniMax-M3`**
**核心能力**：**Coding/Agentic SOTA**、**1M 超长上下文**、**多模态**。
**`MiniMax-M3` 新特性：**
1. 支持图片、视频理解，可参考右方示例代码
2. 支持通过 `thinking` 参数控制思考
## OpenAPI
````yaml api-reference/text/api/openapi-chat-openai.json POST /v1/chat/completions
openapi: 3.1.0
info:
title: MiniMax Text API OpenAI
description: |
MiniMax 文本生成 API，支持对话补全与流式输出。
> ⚡ **`MiniMax-M3` 亮点** —— **Coding/Agentic SOTA**、**1M 超长上下文**、**多模态**。
license:
name: MIT
version: 1.0.0
servers:
- url: https://api.minimaxi.com
security:
- bearerAuth: []
paths:
/v1/chat/completions:
post:
tags:
- Text Generation
summary: Text Generation OpenAI
operationId: chatCompletionOpenAI
parameters:
- name: Content-Type
in: header
required: true
description: 请求体的媒介类型，请设置为 `application/json`，确保请求数据的格式为 JSON
schema:
type: string
enum:
- application/json
default: application/json
requestBody:
content:
application/json:
schema:
$ref: '#/components/schemas/ChatCompletionReq'
examples:
图片理解:
value:
model: MiniMax-M3
thinking:
type: adaptive
messages:
- role: user
content:
- type: text
text: 这张图片的内容是什么？
- type: image_url
image_url:
url: >-
https://filecdn.minimax.chat/public/fe9d04da-f60e-444d-a2e0-18ae743add33.jpeg
max_completion_tokens: 500
视频理解:
value:
model: MiniMax-M3
thinking:
type: adaptive
messages:
- role: user
content:
- type: text
text: 这个视频里发生了什么？
- type: video_url
video_url:
url: >-
https://filecdn.minimax.chat/public/ee8c1648-21f1-41b7-8397-65022d22ffe5.mp4
max_completion_tokens: 500
深度思考:
value:
model: MiniMax-M3
thinking:
type: adaptive
messages:
- role: user
content: 9.11 和 9.9 哪个更大？
max_completion_tokens: 500
流式:
value:
model: MiniMax-M3
thinking:
type: adaptive
messages:
- role: user
content:
- type: text
text: 这张图片的内容是什么？
- type: image_url
image_url:
url: >-
https://filecdn.minimax.chat/public/fe9d04da-f60e-444d-a2e0-18ae743add33.jpeg
stream: true
max_completion_tokens: 500
工具调用:
value:
model: MiniMax-M3
messages:
- role: user
content: 旧金山现在天气怎么样？
tools:
- type: function
function:
name: get_weather
description: Get the current weather for a given location.
parameters:
type: object
properties:
location:
type: string
description: >-
The city and state/country, e.g. San Francisco,
US
required:
- location
required: true
responses:
'200':
description: ''
content:
application/json:
schema:
$ref: '#/components/schemas/ChatCompletionResp'
examples:
图片理解:
value:
id: 066a2a568140d42ba2020cec72d592c0
choices:
- finish_reason: stop
index: 0
message:
content: >-
The user is asking in Chinese what the content of
this image is. Let me describe the image in detail
in Chinese.
这张图片是一张温馨的人像摄影作品，画面内容如下：
**主体人物：**
- 一个可爱的小女孩，大约3-5岁左右
- 她有一头蓬松的棕色卷发，额前有可爱的刘海
- 有着大大的棕绿色眼睛，目光清澈明亮
- 嘴角微微上扬，展露出甜美、纯真的微笑
- 脸颊丰满，皮肤白皙光滑，透着孩童特有的红润
**服装：**
- 身穿一件米白色或奶油色的连衣裙
- 衣领和肩部有精致的蕾丝花边装饰，带有荷叶边设计
- 显得十分优雅可爱
**构图与光线：**
- 这是一张特写肖像照，聚焦于女孩的面部表情
- 采用柔和的暖色调光线，营造出温馨梦幻的氛围
- 背景是模糊的暖棕色调，采用了浅景深（背景虚化）效果
- 整体呈现出油画般的质感，画风柔和、温暖
整张照片充满了童真和纯朴之美，捕捉到了小女孩天真烂漫的瞬间。
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780152150
model: MiniMax-M3
object: chat.completion
usage:
total_tokens: 1604
total_characters: 0
prompt_tokens: 1365
completion_tokens: 239
prompt_tokens_details:
cached_tokens: 114
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
base_resp:
status_code: 0
status_msg: ''
视频理解:
value:
id: 066b4556dd449d59fdfbc602ba47bec9
choices:
- finish_reason: stop
index: 0
message:
content: >-
The user wants me to describe the events in the
provided video.
1. **Analyze the initial state (0.0s - 0.1s):** A
cat is walking on a diving board. The background
shows an Olympic setting, specifically London 2012,
with the rings and "London 2012" text visible. There
are spectators in the stands. A person is standing
at the end of the board.
2. **Observe the action (0.1s - 0.2s):** The cat
continues to walk confidently along the diving
board. It approaches the end of the board. The
person at the end moves out of the way.
3. **Identify the climax (0.2s - 0.3s):** The cat
reaches the end of the diving board, leaps off, and
begins a mid-air rotation. It tucks its body,
performing a dive.
4. **Observe the resolution (0.3s - 0.4s):** The
cat completes its dive, extending its body to enter
the water headfirst. It creates a splash upon entry.
5. **Synthesize the description:**
* Start with the setting: A cat on a diving board at an Olympic venue (London 2012).
* Describe the main action: The cat walks to the end of the board.
* Describe the jump: The cat performs a flip or dive off the board.
* Describe the conclusion: The cat enters the water, creating a splash.
6. **Refine the language:** Make it sound natural
and descriptive. "A cat is seen walking on a diving
board..." "It then proceeds to jump off..."
"...performing a flip before plunging into the water
below." This covers all the key elements observed in
the video clip.
The video shows a cat walking on a diving board. It
then proceeds to jump off, performing a flip before
plunging into the water below.
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780224598
model: MiniMax-M3
object: chat.completion
usage:
total_tokens: 2647
total_characters: 0
prompt_tokens: 2243
completion_tokens: 404
prompt_tokens_details:
cached_tokens: 114
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
base_resp:
status_code: 0
status_msg: ''
深度思考:
value:
id: 066b36619b147e326d17053cccdef70f
choices:
- finish_reason: stop
index: 0
message:
content: >-
The user is asking which is larger: 9.11 or 9.9.
Let me compare these two numbers:
- 9.11
- 9.9 = 9.90
Comparing 9.11 and 9.90:
- The integer parts are both 9.
- Compare the decimal parts: 11 vs 90 (treating as
cents: 0.11 vs 0.90)
- 90 > 11, so 9.9 > 9.11
This is actually a famous question that is sometimes
used to trick people who might confuse string
comparison with numerical comparison (since "9.11" >
"9.9" as strings, but numerically 9.9 > 9.11).
The answer is straightforward: 9.9 is larger.
**9.9 更大。**
比较方法很简单：把小数位数对齐：
- 9.11 = 9.11
- 9.90 = 9.9
整数部分都是 9，所以比较小数部分：90 > 11，因此 **9.9 > 9.11**。
---
这是一个经典的小数比较问题，容易被混淆的原因在于：如果按**字符串**逐字符比较，"9.9"中的 '9'
会大于 "9.11"中的 '1'，从而得出相反的结论。但在**数值**上，9.9 = 9.90 显然大于
9.11。
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780220769
model: MiniMax-M3
object: chat.completion
usage:
total_tokens: 473
total_characters: 0
prompt_tokens: 170
completion_tokens: 303
prompt_tokens_details:
cached_tokens: 157
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
base_resp:
status_code: 0
status_msg: ''
流式:
value:
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
The user
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: ' is asking in Chinese what the content of this image is. Let me describe the image in detail.'
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
这张图片是一个小女孩的肖像特写照片。
**图片内容描述：**
- **
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |
人物**：照片中是一个大约3-5岁的小女孩，她正面面对镜头
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
- **外貌特征**：
- 棕色的波浪卷发，前面有
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
齐刘海，头发似乎在顶部扎了起来
- 大大的浅褐色/
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
榛色眼睛，眼神清澈明亮
- 皮肤白皙，脸颊微微泛红
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |2-
- 嘴角带着温柔、腼腆的微笑
- **服装**：身穿一件白色
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: 或米色的连衣裙，肩部有精致的蕾丝褶皱装饰，领口也
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
带有蕾丝花边
- **背景**：背景是柔和的暖灰色调，营造出温馨
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
的氛围
- **光线**：采用了柔和的暖色调光线，类似于经典肖像画的
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: 打光方式（类似伦勃朗光），光从一侧照在女孩脸上
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
，突出了她的面部轮廓和发丝的光泽
- **整体风格**：照片具有油画般的质感
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- finish_reason: stop
index: 0
delta:
content: ，色彩温暖，是一张非常经典的儿童肖像摄影作品
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
工具调用:
value:
id: 066b13db03518f86c2e3c9b073c04272
choices:
- finish_reason: tool_calls
index: 0
message:
content: >-
The user is asking about the current weather in San
Francisco. I should use the get_weather tool to
fetch this information.
我来帮你查询旧金山的当前天气。
role: assistant
name: MiniMax AI
tool_calls:
- id: call_function_p4iiqtpnh5bj_1
type: function
function:
name: get_weather
arguments: '{"location": "San Francisco, US"}'
index: 0
audio_content: ''
created: 1780211931
model: MiniMax-M3
object: chat.completion
usage:
total_tokens: 477
total_characters: 0
prompt_tokens: 411
completion_tokens: 66
prompt_tokens_details:
cached_tokens: 114
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
base_resp:
status_code: 0
status_msg: ''
text/event-stream:
schema:
$ref: '#/components/schemas/ChatCompletionChunk'
examples:
流式:
value:
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
The user
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: ' is asking in Chinese what the content of this image is. Let me describe the image in detail.'
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
这张图片是一个小女孩的肖像特写照片。
**图片内容描述：**
- **
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |
人物**：照片中是一个大约3-5岁的小女孩，她正面面对镜头
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
- **外貌特征**：
- 棕色的波浪卷发，前面有
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
齐刘海，头发似乎在顶部扎了起来
- 大大的浅褐色/
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
榛色眼睛，眼神清澈明亮
- 皮肤白皙，脸颊微微泛红
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |2-
- 嘴角带着温柔、腼腆的微笑
- **服装**：身穿一件白色
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: 或米色的连衣裙，肩部有精致的蕾丝褶皱装饰，领口也
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
带有蕾丝花边
- **背景**：背景是柔和的暖灰色调，营造出温馨
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
的氛围
- **光线**：采用了柔和的暖色调光线，类似于经典肖像画的
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: 打光方式（类似伦勃朗光），光从一侧照在女孩脸上
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- index: 0
delta:
content: |-
，突出了她的面部轮廓和发丝的光泽
- **整体风格**：照片具有油画般的质感
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
- id: 066a2db7cb70134c45f5d6443d434c2c
choices:
- finish_reason: stop
index: 0
delta:
content: ，色彩温暖，是一张非常经典的儿童肖像摄影作品
role: assistant
name: MiniMax AI
audio_content: ''
created: 1780153015
model: MiniMax-M3
object: chat.completion.chunk
usage: null
input_sensitive: false
output_sensitive: false
input_sensitive_type: 0
output_sensitive_type: 0
output_sensitive_int: 0
components:
schemas:
ChatCompletionReq:
type: object
required:
- model
- messages
properties:
model:
type: string
description: 模型 ID
enum:
- MiniMax-M3
- MiniMax-M2.7
- MiniMax-M2.7-highspeed
- MiniMax-M2.5
- MiniMax-M2.5-highspeed
- MiniMax-M2.1
- MiniMax-M2.1-highspeed
- MiniMax-M2
service_tier:
type: string
description: >-
请求准入服务层级。支持的取值为 `standard` 和 `priority`。省略时默认使用
`standard`。`priority` 的[价格](/guides/pricing-paygo)为 `standard` 的 1.5
倍，并会确保请求获得优先准入，使其排在其他请求之前处理，从而带来更快响应并减少失败。
enum:
- standard
- priority
default: standard
messages:
type: array
description: 包含对话历史的消息列表。支持文本、图片、视频和工具调用。
items:
$ref: '#/components/schemas/Message'
thinking:
type: object
description: >-
控制 MiniMax-M3 thinking。省略时默认开启 adaptive thinking，响应会包含 thinking
内容。对于 M2.x 模型，thinking 无法关闭。
properties:
type:
type: string
enum:
- disabled
- adaptive
default: adaptive
description: >-
thinking 控制类型。
- `disabled`：让 MiniMax-M3 跳过 thinking 并直接回答。对于 M2.x 模型，thinking
仍会保持开启。
- `adaptive`：为 MiniMax-M3 开启 adaptive thinking。省略 `thinking`
时默认使用该值。
required: []
reasoning_split:
type: boolean
description: >-
输出格式开关。启用后将 thinking 内容拆分到 `reasoning_content` 和 `reasoning_details`
字段。这不会开启或关闭 thinking。
stream:
type: boolean
description: 是否使用流式传输，默认为 `false`。设置为 `true` 后，响应将分批返回。
default: false
stream_options:
type: object
description: 流式响应选项。
properties:
include_usage:
type: boolean
default: false
description: 是否在流式响应中包含 token 用量。
max_completion_tokens:
type: integer
format: int64
description: >-
指定生成内容长度的上限（Token 数）。MiniMax-M3 推荐值为 131072（128K），上限为
524288（512K）；其他模型推荐值为 65536（64K），上限为 204800（200K）。如果生成因 `length`
原因中断，请尝试调高此值。
minimum: 1
temperature:
type: number
format: double
description: 温度系数，影响输出随机性，取值范围 [0, 2]，默认值为 1。值越高，输出越随机；值越低，输出越确定。
minimum: 0
maximum: 2
default: 1
top_p:
type: number
format: double
description: 核采样参数，取值范围 [0, 1]。MiniMax-M3 默认值为 0.95，M2.x 系列模型默认值为 0.9。
minimum: 0
maximum: 1
default: 0.95
tools:
type: array
description: 工具定义列表，当前支持 function 工具。
items:
type: object
properties:
type:
type: string
enum:
- function
description: 工具类型，固定为 function。
function:
type: object
properties:
name:
type: string
description: 函数名称。
description:
type: string
description: 函数描述。
parameters:
type: object
description: 函数参数 JSON Schema。
required:
- name
- parameters
required:
- type
- function
max_tokens:
type: integer
format: int64
deprecated: true
description: 旧版生成长度限制参数。已弃用，请改用 `max_completion_tokens`。
minimum: 1
ChatCompletionResp:
type: object
properties:
id:
type: string
description: 本次响应的唯一 ID
choices:
type: array
description: 响应选择列表
items:
type: object
properties:
finish_reason:
type: string
description: 生成停止的原因：`stop`、`length`、`content_filter` 或 `tool_calls`。
enum:
- stop
- length
- content_filter
- tool_calls
index:
type: integer
description: 选项的索引，从 0 开始
message:
type: object
description: 模型生成的完整回复
required:
- content
- role
properties:
content:
type: string
description: 文本回复内容
reasoning_content:
type: string
description: 思考内容。仅在启用 reasoning_split 时返回。
reasoning_details:
type: array
description: 结构化思考内容。仅在启用 reasoning_split 时返回。
items:
type: object
properties:
type:
type: string
description: 思考内容类型，例如 `reasoning.text`。
id:
type: string
description: 思考片段的标识符。
format:
type: string
description: 思考内容的格式标识，例如 `MiniMax-response-v1`。
index:
type: integer
description: 思考片段的顺序索引。
text:
type: string
description: 该片段的思考文本。
role:
type: string
description: 角色，固定为 `assistant`
enum:
- assistant
tool_calls:
type: array
description: >-
模型生成的工具调用列表。仅当 `finish_reason` 为 `tool_calls` 时返回。下一轮请求需在
assistant 消息中原样回带这些调用，并为每个调用追加一条 `role: tool` 消息，其
`tool_call_id` 取自此处的 `id`。
items:
$ref: '#/components/schemas/ToolCall'
created:
type: integer
format: int64
description: 响应创建的 Unix 时间戳（秒）
model:
type: string
description: 本次请求使用的模型 ID
object:
type: string
description: 对象类型。非流式为 `chat.completion`，流式为 `chat.completion.chunk`
enum:
- chat.completion
- chat.completion.chunk
usage:
$ref: '#/components/schemas/Usage'
input_sensitive:
type: boolean
description: 输入内容是否命中敏感词。如果输入内容严重违规，接口会返回内容违规错误信息，回复内容为空
input_sensitive_type:
type: integer
format: int64
description: >-
输入命中敏感词类型，当input_sensitive为true时返回。取值为以下其一：1 严重违规；2 色情；3 广告；4 违禁；5
谩骂；6 暴恐；7 其他
output_sensitive:
type: boolean
description: 输出内容是否命中敏感词。如果输出内容严重违规，接口会返回内容违规错误信息，回复内容为空
output_sensitive_type:
type: integer
format: int64
description: 输出命中敏感词类型
base_resp:
type: object
description: 错误状态码和详情
properties:
status_code:
type: integer
format: int64
description: |-
状态码
- `1000`: 未知错误
- `1001`: 请求超时
- `1002`: 触发限流
- `1004`: 鉴权失败
- `1008`: 余额不足
- `1013`: 服务内部错误
- `1027`: 输出内容错误
- `1039`: Token 超出限制
- `2013`: 参数错误
更多内容可查看 [错误码查询列表](/api-reference/errorcode) 了解详情
status_msg:
type: string
description: 错误详情
ChatCompletionChunk:
type: object
description: ''
properties:
id:
type: string
description: 本次响应的唯一 ID
choices:
type: array
description: 流式响应选择列表
items:
type: object
properties:
index:
type: integer
description: 选项的索引，从 0 开始
delta:
type: object
description: 增量内容
properties:
role:
type: string
description: 角色，固定为 `assistant`
enum:
- assistant
content:
type: string
description: 增量文本内容
finish_reason:
type: string
nullable: true
description: >-
生成停止的原因，未结束时为 null：`stop`（自然结束），`length`（达到
`max_completion_tokens` 上限）
enum:
- stop
- length
created:
type: integer
format: int64
description: 响应创建的 Unix 时间戳（秒）
model:
type: string
description: 本次请求使用的模型 ID
object:
type: string
description: 对象类型，固定为 `chat.completion.chunk`
enum:
- chat.completion.chunk
usage:
$ref: '#/components/schemas/Usage'
description: Token 使用情况（仅在最后一个 chunk 中返回）
input_sensitive_type:
type: integer
format: int64
description: 输入命中敏感词类型
output_sensitive:
type: boolean
description: 输出内容是否命中敏感词
output_sensitive_type:
type: integer
format: int64
description: 输出命中敏感词类型
Message:
type: object
required:
- role
properties:
role:
type: string
enum:
- system
- user
- assistant
- tool
description: 消息发送者的角色。
name:
type: string
description: 发送者的名称。若同一类型的角色有多个，须提供具体名称以区分
content:
oneOf:
- type: string
description: 文本消息内容
- type: array
description: >-
多模态消息内容块。MiniMax-M3 支持文本、图片和视频输入。
**多模态文件大小及容量限制**
- URL / Base64 方式传入：视频文件 ≤ 50 MB，图片文件 ≤ 10 MB，请求体 ≤ 64 MB。
- 使用 Files API 上传：在 `content` 中以 `mm_file://{file_id}`
形式引用，单个视频最大 512 MB。
items:
$ref: '#/components/schemas/MessageContentPart'
description: 消息内容。MiniMax-M3 支持文本、图片和视频内容块。
tool_calls:
type: array
description: assistant 消息中的工具调用列表。
items:
$ref: '#/components/schemas/ToolCall'
tool_call_id:
type: string
description: >-
工具调用 ID。当 `role` 为 `tool` 时必填，取值为上一轮 assistant 消息中 `tool_calls`
内对应一项的 `id`，用于将本条工具返回结果与具体的工具调用关联。其他角色下该字段会被忽略。
ToolCall:
type: object
required:
- id
- type
- function
description: assistant 消息中模型生成的工具调用。
properties:
id:
type: string
description: 工具调用 ID，由模型生成。
type:
type: string
enum:
- function
description: 工具调用类型。当前仅支持 `function`。
function:
type: object
required:
- name
- arguments
properties:
name:
type: string
description: 待调用函数名称。
arguments:
type: string
description: 函数入参，JSON 格式字符串。
Usage:
type: object
description: 本次请求的 Token 使用情况统计
properties:
total_tokens:
type: integer
description: 消耗的总 Token 数
MessageContentPart:
type: object
required:
- type
description: 多模态消息内容块。每个内容块通过 `type` 声明类型，仅填写与该类型匹配的字段。
properties:
type:
type: string
enum:
- text
- image_url
- video_url
description: |-
内容块类型：
- `text`：文本块
- `image_url`：图片输入
- `video_url`：视频输入
text:
type: string
description: 文本内容（当 `type` 为 `text` 时）。
image_url:
type: object
description: |-
图片输入（当 `type` 为 `image_url` 时）。单张图片最大 10 MB。
**支持的图片格式**
| 格式 | 常见拓展名 | MIME Type |
| :-- | :-- | :-- |
| JPEG | .jpg, .jpeg | image/jpeg |
| PNG | .png | image/png |
| GIF | .gif | image/gif |
| WEBP | .webp | image/webp |
required:
- url
properties:
url:
type: string
description: 图片 URL 或 Base64 data URL。
detail:
type: string
enum:
- low
- default
- high
default: default
description: |-
控制图片解析分辨率，默认值为 default。
单张图片粗略 token 用量估算：
| detail | 粗略 token 用量 |
| :-- | :-- |
| low | 通常为几百 token，最高约 600 |
| default | 通常约 1k-3k token，最高约 5k |
| high | 通常为数千 token，最高约 15k+ |
实际用量取决于图片尺寸和内容；请以响应中的 usage 或可用的 token 计数接口为准。
max_long_side_pixel:
type: integer
minimum: 1
description: 图片最长边像素限制。
video_url:
type: object
description: >-
视频输入（当 `type` 为 `video_url` 时）。
**支持的视频格式**
| 格式 | 常见拓展名 | MIME Type |
| :-- | :-- | :-- |
| MP4 | .mp4 | video/mp4 |
| AVI | .avi | video/avi 或 video/x-msvideo |
| MOV | .mov | url 传入视频：对象存储请设置 Content-Type 为
video/quicktime；base64 编码：请使用 video/mov，即
data:video/mov;base64, |
| MKV | .mkv | video/x-matroska |
required:
- url
properties:
url:
type: string
description: >-
视频 URL、Base64 data URL，或形如 `mm_file://{file_id}` 的 Files API
引用。`file_id`
需先通过[文件上传](/api-reference/file-management-upload)接口上传视频后获取。URL 或
Base64 视频最大 50 MB；Files API 视频最大 512 MB。
detail:
type: string
enum:
- low
- default
- high
default: default
description: 控制视频抽帧分辨率，默认值为 default。
fps:
type: number
minimum: 0.2
maximum: 5
default: 1
description: |-
视频抽帧频率。默认值为 1，范围 [0.2, 5]。
- 取值越高：对画面变化越敏感，token 花费高、速度慢。
- 取值越低：token 花费少、速度快，但对画面变化迟钝。
max_long_side_pixel:
type: integer
minimum: 1
description: 视频单帧最长边像素限制。
securitySchemes:
bearerAuth:
type: http
scheme: bearer
bearerFormat: JWT
description: |-
`HTTP: Bearer Auth`
- Security Scheme Type: http
- HTTP Authorization Scheme: Bearer API_key，用于验证账户信息，可在 [账户管理>接口密钥](https://platform.minimaxi.com/user-center/basic-information/interface-key) 中查看
````