Messages Protocol

Access Anthropic, DeepSeek, DashScope, Xiaomi, Kimi, MiniMax, and Zai models through the Anthropic Messages API format.

The Messages protocol exposes the Anthropic Messages API format at POST /v1/messages. It is the most broadly supported protocol on OpenModel, serving seven providers: Anthropic, DeepSeek, DashScope, Xiaomi, Kimi, MiniMax, and Zai.

Because every provider shares the same endpoint and request shape, you use the Anthropic SDK for all of them — switching providers means changing only the model parameter, with no other code changes. See the Anthropic SDK guide for installation and full examples.

Endpoint

POST https://api.openmodel.ai/v1/messages

A second endpoint, POST /v1/messages/count_tokens, counts tokens without creating a message (Anthropic only).

Authentication

All providers authenticate with a single OpenModel API key. Pass it either as a bearer token or via the X-Api-Key header:

Authorization: Bearer om-your-api-key

The anthropic-version header is accepted for SDK compatibility and defaults to 2023-06-01.

Switching providers

To target a different provider, change only the model parameter:

# DeepSeek
client.messages.create(model="deepseek-v4-flash", max_tokens=1024, messages=[...])
# DashScope
client.messages.create(model="qwen3-max", max_tokens=1024, messages=[...])
# Xiaomi
client.messages.create(model="mimo-v2.5-pro", max_tokens=1024, messages=[...])
# Kimi
client.messages.create(model="kimi-k2.5", max_tokens=1024, messages=[...])
# MiniMax
client.messages.create(model="MiniMax-M2.5", max_tokens=1024, messages=[...])
# Zai
client.messages.create(model="glm-5", max_tokens=1024, messages=[...])

Provider differences & playground

Select a provider below to see its example models, error-response format, and an interactive playground pre-filled with that provider's model. Most providers behave identically to the Anthropic baseline; only the documented differences vary.

Example models: claude-opus-4-6claude-opus-4-7claude-opus-4-8claude-sonnet-4-6claude-haiku-4-5-20251001
Error response format: Standard Anthropic format — errors carry a string error.type (e.g. authentication_error, rate_limit_error). This is the baseline all Messages providers are compared against.
Official documentation: Anthropic Messages API

These are example models; the available list may change. View all available models in the Console

Authorization

AuthorizationBearer <token>

Include your API key: Authorization: Bearer om-your-api-key

In: header

Header Parameters

anthropic-version*string

Required. Specifies the Anthropic API version.

Default"2023-06-01"

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

model*string

The model ID to use (e.g. claude-sonnet-4-6, claude-opus-4-8).

messages*array<>

An array of message objects representing the conversation.

max_tokens*integer

The maximum number of tokens to generate before stopping.

system?|array<>

System prompt that sets the context and behavior for the assistant.

stream?boolean

Whether to stream the response using server-sent events (SSE).

Defaultfalse

temperature?number

Amount of randomness in the response. Default is 1.0.

Range0 <= value <= 1

top_p?number

Nucleus sampling. Use a value between 0 and 1.

top_k?integer

Only sample from the top K options for each subsequent token.

tools?array<>

Definitions of tools that the model may use.

tool_choice?

Controls how the model uses the provided tools.

metadata?

An object describing metadata about the request.

stop_sequences?array<string>

Custom text sequences that will cause the model to stop generating.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "claude-sonnet-4-6",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "Explain the concept of recursion in programming."      }    ]  }'

{
  "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Recursion is a programming technique where a function calls itself to solve a problem by breaking it down into smaller, identical subproblems."
    }
  ],
  "model": "claude-sonnet-4-6",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 15,
    "output_tokens": 168
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

Authorization

AuthorizationBearer <token>

Include your API key: Authorization: Bearer om-your-api-key

In: header

Header Parameters

anthropic-version*string

Required. Specifies the Anthropic API version.

Default"2023-06-01"

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

model*string

The model ID to use for tokenization. Different models may tokenize text differently.

messages*array<>

The messages array to count tokens for. Uses the same format as the Messages API.

system?|array<>

System prompt to include in the token count.

tools?array<>

Tool definitions to include in the token count.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages/count_tokens" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "claude-sonnet-4-6",    "messages": [      {        "role": "user",        "content": "What is the meaning of life?"      }    ]  }'

{
  "input_tokens": 15
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

Example models: deepseek-v4-flashdeepseek-v4-pro
Error response format: Standard Anthropic error.type format.

These are example models; the available list may change. View all available models in the Console

Compatibility notes

DeepSeek's Anthropic-format endpoint does not implement every Anthropic feature. Behavior differs from Anthropic as follows.

Field	Status
`max_tokens`, `system`, `stream`, `stop_sequences`, `top_p`, `tools`	Fully supported
`temperature`	Supported, range `0.0 – 2.0`
`tool_choice`	Supported (`none` / `auto` / `any` / `tool`); `disable_parallel_tool_use` ignored
`thinking`	Supported (`budget_tokens` ignored)
`top_k`	Ignored
`metadata`	Ignored
`cache_control` (on `tools` and message blocks)	Ignored
`anthropic-beta`, `anthropic-version` headers	Ignored
`tool_result.is_error`	Ignored

Unsupported content block types: image, document, search_result, redacted_thinking, server_tool_use, web_search_tool_result, code_execution_tool_result, mcp_tool_use, mcp_tool_result, container_upload.

Model fallback: passing an unsupported model name is silently mapped to deepseek-v4-flash by the DeepSeek backend.

Authorization

AuthorizationBearer <token>

Include your API key: Authorization: Bearer om-your-api-key

In: header

Header Parameters

anthropic-version*string

Required. Specifies the Anthropic API version.

Default"2023-06-01"

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

model*string

The DeepSeek model ID to use (e.g. deepseek-v4-flash, deepseek-v4-pro).

messages*array<>

An array of message objects representing the conversation.

max_tokens*integer

The maximum number of tokens to generate before stopping.

system?|array<>

System prompt that sets the context and behavior for the assistant.

stream?boolean

Whether to stream the response using server-sent events (SSE).

Defaultfalse

temperature?number

Amount of randomness in the response. Range 0.0 to 2.0. Default is 1.0.

Range0 <= value <= 2

top_p?number

Nucleus sampling. Use a value between 0 and 1.

top_k?integer

Only sample from the top K options for each subsequent token. Ignored by DeepSeek (accepted for Anthropic API compatibility).

tools?array<>

Definitions of tools that the model may use.

tool_choice?

Controls how the model uses the provided tools. disable_parallel_tool_use is ignored by DeepSeek.

metadata?

An object describing metadata about the request. Ignored by DeepSeek (accepted for Anthropic API compatibility).

stop_sequences?array<string>

Custom text sequences that will cause the model to stop generating.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "deepseek-v4-flash",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "Explain the concept of recursion in programming."      }    ]  }'

{
  "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Recursion is a programming technique where a function calls itself to solve a problem by breaking it down into smaller, identical subproblems."
    }
  ],
  "model": "deepseek-v4-flash",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 15,
    "output_tokens": 168
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

Example models: qwen3.7-maxqwen3.6-plusqwen3.6-max-previewqwen3.6-flashqwen3-max
Error response format: The Messages endpoint returns DashScope's native error format — a root-level code/type with no error wrapper. The Responses endpoint uses the standard error.code + error.type shape.
Official documentation: DashScope Anthropic API

These are example models; the available list may change. View all available models in the Console

Authorization

AuthorizationBearer <token>

Include your API key: Authorization: Bearer om-your-api-key

In: header

Header Parameters

anthropic-version*string

Required. Specifies the Anthropic API version.

Default"2023-06-01"

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

model*string

The DashScope model ID to use (e.g. qwen3-max).

messages*array<>

An array of message objects representing the conversation.

max_tokens*integer

The maximum number of tokens to generate before stopping.

system?|array<>

System prompt that sets the context and behavior for the assistant.

stream?boolean

Whether to stream the response using server-sent events (SSE).

Defaultfalse

temperature?number

Amount of randomness in the response. Default is 1.0.

Range0 <= value <= 1

top_p?number

Nucleus sampling. Use a value between 0 and 1.

top_k?integer

Only sample from the top K options for each subsequent token.

tools?array<>

Definitions of tools that the model may use.

tool_choice?

Controls how the model uses the provided tools.

metadata?

An object describing metadata about the request.

stop_sequences?array<string>

Custom text sequences that will cause the model to stop generating.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "qwen3-max",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "Explain the concept of recursion in programming."      }    ]  }'

{
  "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Recursion is a programming technique where a function calls itself to solve a problem by breaking it down into smaller, identical subproblems."
    }
  ],
  "model": "qwen3-max",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 15,
    "output_tokens": 168
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

Example models: mimo-v2-flashmimo-v2-omnimimo-v2-promimo-v2.5mimo-v2.5-pro
Error response format: Anthropic-compatible error.type format. The gateway additionally retries on 529 (API overloaded) responses.
Official documentation: Xiaomi MiMo Anthropic API

These are example models; the available list may change. View all available models in the Console

Authorization

AuthorizationBearer <token>

Include your API key: Authorization: Bearer om-your-api-key

In: header

Header Parameters

anthropic-version*string

Required. Specifies the Anthropic API version.

Default"2023-06-01"

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

model*string

The Xiaomi MiMo model ID to use (e.g. mimo-v2.5-pro, mimo-v2-pro, mimo-v2.5, mimo-v2-omni, mimo-v2-flash).

messages*array<>

An array of message objects representing the conversation.

max_tokens*integer

The maximum number of tokens to generate before stopping.

system?|array<>

System prompt that sets the context and behavior for the assistant.

stream?boolean

Whether to stream the response using server-sent events (SSE).

Defaultfalse

temperature?number

Amount of randomness in the response. Default is 1.0.

Range0 <= value <= 1

top_p?number

Nucleus sampling. Use a value between 0 and 1.

top_k?integer

Only sample from the top K options for each subsequent token.

tools?array<>

Definitions of tools that the model may use.

tool_choice?

Controls how the model uses the provided tools.

metadata?

An object describing metadata about the request.

stop_sequences?array<string>

Custom text sequences that will cause the model to stop generating.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "mimo-v2.5-pro",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "Explain the concept of recursion in programming."      }    ]  }'

{
  "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Recursion is a programming technique where a function calls itself to solve a problem by breaking it down into smaller, identical subproblems."
    }
  ],
  "model": "mimo-v2.5-pro",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 15,
    "output_tokens": 168
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

Example models: kimi-k2.5kimi-k2.6
Error response format: Standard Anthropic error.type format.
Official documentation: Kimi Platform

These are example models; the available list may change. View all available models in the Console

Authorization

AuthorizationBearer <token>

Include your API key: Authorization: Bearer om-your-api-key

In: header

Header Parameters

anthropic-version*string

Required. Specifies the Anthropic API version.

Default"2023-06-01"

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

model*string

The model ID to use (e.g. claude-sonnet-4-6, claude-opus-4-8).

messages*array<>

An array of message objects representing the conversation.

max_tokens*integer

The maximum number of tokens to generate before stopping.

system?|array<>

System prompt that sets the context and behavior for the assistant.

stream?boolean

Whether to stream the response using server-sent events (SSE).

Defaultfalse

temperature?number

Amount of randomness in the response. Default is 1.0.

Range0 <= value <= 1

top_p?number

Nucleus sampling. Use a value between 0 and 1.

top_k?integer

Only sample from the top K options for each subsequent token.

tools?array<>

Definitions of tools that the model may use.

tool_choice?

Controls how the model uses the provided tools.

metadata?

An object describing metadata about the request.

stop_sequences?array<string>

Custom text sequences that will cause the model to stop generating.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "claude-sonnet-4-6",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "Explain the concept of recursion in programming."      }    ]  }'

{
  "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Recursion is a programming technique where a function calls itself to solve a problem by breaking it down into smaller, identical subproblems."
    }
  ],
  "model": "claude-sonnet-4-6",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 15,
    "output_tokens": 168
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

Example models: MiniMax-M3MiniMax-M2.7MiniMax-M2.5
Error response format: Standard Anthropic error.type format.
Official documentation: MiniMax Platform

These are example models; the available list may change. View all available models in the Console

Authorization

AuthorizationBearer <token>

Include your API key: Authorization: Bearer om-your-api-key

In: header

Header Parameters

anthropic-version*string

Required. Specifies the Anthropic API version.

Default"2023-06-01"

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

model*string

The model ID to use (e.g. claude-sonnet-4-6, claude-opus-4-8).

messages*array<>

An array of message objects representing the conversation.

max_tokens*integer

The maximum number of tokens to generate before stopping.

system?|array<>

System prompt that sets the context and behavior for the assistant.

stream?boolean

Whether to stream the response using server-sent events (SSE).

Defaultfalse

temperature?number

Amount of randomness in the response. Default is 1.0.

Range0 <= value <= 1

top_p?number

Nucleus sampling. Use a value between 0 and 1.

top_k?integer

Only sample from the top K options for each subsequent token.

tools?array<>

Definitions of tools that the model may use.

tool_choice?

Controls how the model uses the provided tools.

metadata?

An object describing metadata about the request.

stop_sequences?array<string>

Custom text sequences that will cause the model to stop generating.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "claude-sonnet-4-6",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "Explain the concept of recursion in programming."      }    ]  }'

{
  "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Recursion is a programming technique where a function calls itself to solve a problem by breaking it down into smaller, identical subproblems."
    }
  ],
  "model": "claude-sonnet-4-6",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 15,
    "output_tokens": 168
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

Example models: glm-5.1glm-5glm-4.7
Error response format: Standard Anthropic error.type format.
Official documentation: Zai Platform

These are example models; the available list may change. View all available models in the Console

Authorization

AuthorizationBearer <token>

Include your API key: Authorization: Bearer om-your-api-key

In: header

Header Parameters

anthropic-version*string

Required. Specifies the Anthropic API version.

Default"2023-06-01"

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

model*string

The model ID to use (e.g. claude-sonnet-4-6, claude-opus-4-8).

messages*array<>

An array of message objects representing the conversation.

max_tokens*integer

The maximum number of tokens to generate before stopping.

system?|array<>

System prompt that sets the context and behavior for the assistant.

stream?boolean

Whether to stream the response using server-sent events (SSE).

Defaultfalse

temperature?number

Amount of randomness in the response. Default is 1.0.

Range0 <= value <= 1

top_p?number

Nucleus sampling. Use a value between 0 and 1.

top_k?integer

Only sample from the top K options for each subsequent token.

tools?array<>

Definitions of tools that the model may use.

tool_choice?

Controls how the model uses the provided tools.

metadata?

An object describing metadata about the request.

stop_sequences?array<string>

Custom text sequences that will cause the model to stop generating.

Response Body

`application/json`

curl -X POST "https://example.com/v1/messages" \  -H "anthropic-version: 2023-06-01" \  -H "Content-Type: application/json" \  -d '{    "model": "claude-sonnet-4-6",    "max_tokens": 1024,    "messages": [      {        "role": "user",        "content": "Explain the concept of recursion in programming."      }    ]  }'

{
  "id": "msg_01XFDUDYJgAACzvnptvVoYEL",
  "type": "message",
  "role": "assistant",
  "content": [
    {
      "type": "text",
      "text": "Recursion is a programming technique where a function calls itself to solve a problem by breaking it down into smaller, identical subproblems."
    }
  ],
  "model": "claude-sonnet-4-6",
  "stop_reason": "end_turn",
  "usage": {
    "input_tokens": 15,
    "output_tokens": 168
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

{
  "type": "string",
  "error": {
    "type": "string",
    "message": "string"
  }
}

Messages Protocol

200application/json

400application/json

401application/json

429application/json

200application/json

400application/json

401application/json

200application/json

400application/json

401application/json

429application/json

200application/json

400application/json

401application/json

429application/json

200application/json

400application/json

401application/json

429application/json

200application/json

400application/json

401application/json

429application/json

200application/json

400application/json

401application/json

429application/json

200application/json

400application/json

401application/json

429application/json

On this page

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`