Create Message

curl --request POST \
  --url https://api.minimax.io/anthropic/v1/messages \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: <content-type>' \
  --data '
{
  "model": "MiniMax-M3.1-Flash",
  "thinking": {
    "type": "adaptive"
  },
  "messages": [
    {
      "role": "user",
      "content": "Which is bigger, 9.11 or 9.9?"
    }
  ],
  "max_tokens": 4096,
  "output_config": {
    "effort": "max"
  }
}
'

{
  "id": "066b367ba949a58a9d6829b4bf8cbcd4",
  "type": "message",
  "role": "assistant",
  "model": "MiniMax-M3.1-Flash",
  "content": [
    {
      "thinking": "The user is asking which is bigger, 9.11 or 9.9.\n\n9.11 vs 9.9\n\n9.9 is greater than 9.11.\n\nTo compare: 9.11 = 9.11 and 9.9 = 9.90\n\n9.90 > 9.11, so 9.9 is bigger.",
      "signature": "6d0315c818f9664ff185dabaa22cd89f2bf28a3a52122095bce23d905471ec5f",
      "type": "thinking"
    },
    {
      "text": "**9.9 is bigger than 9.11.**\n\nTo compare decimals, it helps to write them with the same number of decimal places:\n- 9.11 = 9.11\n- 9.9 = 9.90\n\nSince 9.90 > 9.11, **9.9 is larger**.",
      "type": "text"
    }
  ],
  "usage": {
    "input_tokens": 13,
    "output_tokens": 149,
    "cache_creation_input_tokens": 0,
    "cache_read_input_tokens": 159
  },
  "stop_reason": "end_turn"
}

POST

anthropic

messages

curl --request POST \
  --url https://api.minimax.io/anthropic/v1/messages \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: <content-type>' \
  --data '
{
  "model": "MiniMax-M3.1-Flash",
  "thinking": {
    "type": "adaptive"
  },
  "messages": [
    {
      "role": "user",
      "content": "Which is bigger, 9.11 or 9.9?"
    }
  ],
  "max_tokens": 4096,
  "output_config": {
    "effort": "max"
  }
}
'

{
  "id": "066b367ba949a58a9d6829b4bf8cbcd4",
  "type": "message",
  "role": "assistant",
  "model": "MiniMax-M3.1-Flash",
  "content": [
    {
      "thinking": "The user is asking which is bigger, 9.11 or 9.9.\n\n9.11 vs 9.9\n\n9.9 is greater than 9.11.\n\nTo compare: 9.11 = 9.11 and 9.9 = 9.90\n\n9.90 > 9.11, so 9.9 is bigger.",
      "signature": "6d0315c818f9664ff185dabaa22cd89f2bf28a3a52122095bce23d905471ec5f",
      "type": "thinking"
    },
    {
      "text": "**9.9 is bigger than 9.11.**\n\nTo compare decimals, it helps to write them with the same number of decimal places:\n- 9.11 = 9.11\n- 9.9 = 9.90\n\nSince 9.90 > 9.11, **9.9 is larger**.",
      "type": "text"
    }
  ],
  "usage": {
    "input_tokens": 13,
    "output_tokens": 149,
    "cache_creation_input_tokens": 0,
    "cache_read_input_tokens": 159
  },
  "stop_reason": "end_turn"
}

✨ New model — MiniMax-M3.1-FlashCore capabilities: 1M long context, multimodal, tunable thinking depth.

What’s new in MiniMax-M3.1-Flash:

Image and video understanding — see the example code on the right
Tune thinking depth via output_config.effort, which accepts low, medium, high, xhigh, and max; the default is max
Thinking is on by default with no configuration needed; thinking content is returned separately in thinking content blocks

Authorizations

Authorization

string

header

required

Bearer API Key auth. Send Authorization: Bearer <API_KEY>. If Authorization and x-api-key are both present, Authorization takes precedence.

Headers

Content-Type

enum<string>

default:application/json

required

Media type of the request body, should be set to application/json to ensure JSON format

Available options:

application/json

Body

application/json

model

enum<string>

required

Model ID. MiniMax-M3.1-Flash and MiniMax-M3 are multimodal models with native support for text, image, and video input, alongside tool use and thinking content blocks. MiniMax-M3.1-Flash additionally always thinks and supports tunable thinking depth via output_config.effort.

Available options:

MiniMax-M3.1-Flash,

MiniMax-M3

messages

object[]

required

Conversation history. MiniMax-M3.1-Flash and MiniMax-M3 support text, image, video, tool use, tool result, and thinking content blocks. The M2.7, M2.5, M2.1, and M2 series support text and tool-call content blocks only; they do not support image or video input.

Show child attributes

service_tier

enum<string>

default:standard

Service tier for request admission. Supported values are standard and priority. If omitted, the request uses the standard tier. The priority price is 1.5 times the standard price and ensures priority admission so the request is processed ahead of other requests, leading to faster responses and fewer failures. MiniMax-M3.1-Flash does not support this field.

Available options:

standard,

priority

system

Set the role and behavior of the model.

stream

boolean

default:false

Whether to use streaming output, defaults to false. When set to true, the response will be returned in chunks

max_tokens

integer<int64>

Specifies the upper limit for generated content length (in tokens). For MiniMax-M3.1-Flash and MiniMax-M3 the recommended value is 131072 (128K) and the maximum is 524288 (512K); for other models the recommended value is 65536 (64K) and the maximum is 204800 (200K). Content exceeding the limit will be truncated. If generation stops due to length, try increasing this value

Required range: x >= 1

temperature

number<double>

default:1

Not supported yet; any value passed has no effect. MiniMax-M3.1-Flash and MiniMax-M3 always use 1.

top_p

number<double>

default:0.95

Not supported yet; any value passed has no effect. MiniMax-M3.1-Flash and MiniMax-M3 always use 0.95. If passed, the value must be within (0, 1].

Required range: 0 <= x <= 1

tools

object[]

Tool definitions for Anthropic-compatible tool use.

Show child attributes

tool_choice

object

Tool selection strategy. Only auto is supported; the model decides whether to call tools.

Show child attributes

thinking

object

Controls thinking behavior. The default depends on the model, so no schema-level default is declared.

MiniMax-M3.1-Flash: always thinks. If type is sent it must be adaptive; disabled returns HTTP 400. Use output_config.effort to tune thinking depth.
MiniMax-M3: thinking is off when thinking is omitted; send adaptive to enable it and return thinking blocks.
M2.x models: thinking cannot be disabled; disabled is accepted but ignored.

Show child attributes

output_config

object

Output configuration. Use effort to tune thinking depth for MiniMax-M3.1-Flash.

Show child attributes

metadata

object

Request metadata. user_id is recommended for end-user-level aggregation, rate limiting, and billing analysis.

Show child attributes

Response

string

Unique ID of this response

type

enum<string>

Object type, fixed as message

Available options:

message

role

enum<string>

Role, fixed as assistant

Available options:

assistant

model

string

Model ID used for this request

content

object[]

List of response content blocks

Show child attributes

stop_reason

enum<string>

Reason for stopping generation:

end_turn: Model ended naturally
max_tokens: Reached max_tokens limit
tool_use: Model requested tool use

Available options:

end_turn,

max_tokens,

tool_use

usage

object

Token usage for this request, including prompt cache usage when applicable.

Show child attributes

Using the API

Text

Video

Speech

Image

Music

File

Models

Authorizations

Headers

Body

Response