Text model · Moonshot AI

Kimi K3.1 Pro API

Kimi K3.1 Pro is the Pro build of Moonshot's K3 line. It reasons before every answer, accepts text and images, and works in all three API formats on TokenRa. The reasoning comes back in reasoning, not in reasoning_content.

Model ID: kimi-k3.1-pro
API formats: Chat Completions, Responses, Messages
Input: Text, image
Output: Text

Capabilities

Capabilities & endpoints

Chat generation through three API formats. Each was called successfully on tokenra.io on 2026-10-23; pick one to switch the code examples.

POSThttps://tokenra.io/v1/chat/completionsTested

OpenAI SDK with base_url="https://tokenra.io/v1". Auth: Authorization: Bearer. Recommended.

POSThttps://tokenra.io/v1/responsesTested

OpenAI Responses API, same SDK and key. Text answer in output_text.

POSThttps://tokenra.io/v1/messagesTested

Anthropic SDK or Claude Code with base URL https://tokenra.io. Auth: x-api-key (or Bearer) plus anthropic-version.

Request

Request parameters

Fields for POST /v1/chat/completions. Everything listed was accepted by the model in testing.

Parameter	Type	Required	Description
`model`	string	Yes	The model ID to call, for example `kimi-k3.1-pro`.
`messages`	array	Yes	Conversation messages in order (system, user, assistant, tool). For multi-turn chat, send the earlier turns back in this array. `content` can be a string or an array of `text` and `image_url` parts.
`stream`	boolean	No	Return the answer as server-sent events. Default `false`.
`stream_options`	object	No	With `{"include_usage": true}`, the last chunk carries `usage` and an empty `choices` array.
`max_tokens`	integer	No	Upper limit for generated tokens. Reasoning tokens count toward it: if it is too low the model can spend it all on reasoning and return no text (`finish_reason: "length"`).
`max_completion_tokens`	integer	No	Upper limit for all generated tokens, reasoning included. If it is too low the answer can be empty (`finish_reason: "length"`).
`temperature`	number	No	Sampling temperature.
`top_p`	number	No	Nucleus sampling threshold. Usually tune either this or temperature.
`presence_penalty`	number	No	Penalises tokens that already appeared, nudging the model toward new topics.
`frequency_penalty`	number	No	Penalises frequent tokens to reduce repetition.
`response_format`	object	No	`{"type": "json_object"}` returns a JSON object. `{"type": "json_schema", "json_schema": {...}}` returns JSON that matches your schema (tested with `strict: true`).
`tools`	array	No	Function tools the model may call (`type: "function"`).
`tool_choice`	string / object	No	`auto`, `none`, `required`, or a specific function.
`parallel_tool_calls`	boolean	No	Allow several tool calls in one response.
`reasoning_effort`	string	No	Reasoning depth such as `low`, `medium` or `high`. The model accepted it in testing; how much it changes the reasoning varies by model.
`user`	string	No	Your own end-user identifier for auditing or abuse handling. It is not part of the conversation.

Not supported by this model

Thinking cannot be turned off. thinking: {"type": "disabled"} and enable_thinking: false were both accepted, and the reasoning tokens stayed at 66–71.

Other formats

Format	Required fields	Answer location
OpenAI Responses	`model`, `input`	`output_text`
Anthropic Messages	`model`, `max_tokens`, `messages`	`content[]` blocks of type `text`

Response

Response format

{
  "id": "chatcmpl-1792766122484",
  "object": "chat.completion",
  "created": 1792766122,
  "model": "kimi-k3.1-pro",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "An API gateway is the single entry point that receives client requests and routes them to the right backend service. It also handles shared concerns such as authentication, rate limiting and logging.",
        "reasoning": "The user wants two sentences about API gateways...",
        "reasoning_details": [
          {
            "type": "reasoning.text",
            "text": "The user wants two sentences about API gateways...",
            "format": "unknown",
            "index": 0
          }
        ]
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 26,
    "completion_tokens": 58,
    "total_tokens": 84,
    "prompt_tokens_details": {
      "cached_tokens": 0
    },
    "completion_tokens_details": {
      "reasoning_tokens": 21
    }
  }
}

Field	Type	Description
`id`	string	Completion ID. Include it when you contact support.
`object`	string	`chat.completion`, or `chat.completion.chunk` when streaming.
`created`	integer	Unix timestamp in seconds.
`model`	string	The model that produced the answer.
`choices[].message.content`	string	The answer text.
`choices[].message.reasoning`	string	Some models return the reasoning under `reasoning` instead of `reasoning_content`, and repeat the same text in `reasoning_details`.
`choices[].message.tool_calls`	array	Tool calls requested by the model, when `tools` were sent.
`choices[].finish_reason`	string	`stop`, `length` or `tool_calls`.
`usage`	object	`prompt_tokens`, `completion_tokens`, `total_tokens`, plus `prompt_tokens_details.cached_tokens` and `completion_tokens_details.reasoning_tokens`.

With stream: true the answer arrives as data: lines of chat.completion.chunk objects and ends with data: [DONE].

Errors

Errors & limits

Common errors

HTTP	error.code	Meaning	What to do
`401`	—	Missing, wrong or deleted API key. The message is `Invalid token`.	Check the `Authorization: Bearer` header and the key in the console.
`400`	—	The body is not valid JSON.	Fix the JSON and send `Content-Type: application/json`.
`500`	`invalid_request`	A required field is missing, for example `messages`.	Add the field named in the message. Retrying will not help.
`403`	`insufficient_user_quota`	Your balance or plan quota is used up.	Top up in the console, then retry.
`503`	`model_not_found`	Unknown model ID, or no channel serves it for your group.	Check the ID in the model list or with `GET /v1/models`.
`429`	—	Too many requests.	Back off exponentially and retry.
`5xx`	—	The upstream provider failed or timed out.	Retry a few times; keep the request id if it persists.

Limits and notes

Image input was tested with an image_url part that points to a public HTTPS image.
Reasoning is returned in reasoning (a string) with the same text in reasoning_details, and the answer in content. Read reasoning, not reasoning_content, which this model leaves empty.
Responses took 6.6–19.6 seconds in testing, the image question the longest. Use a client timeout of at least 120 seconds, or stream.
reasoning_effort low, high and max were all accepted, but the depth did not scale with the level on a short question (61, 4 and 132 reasoning tokens), so treat it as a hint rather than a hard control.
Unlike kimi-k3, this model accepted top_p: 0.9 and non-zero penalties without an error.

Capabilities & endpoints

Request parameters

Other formats

Response format

Errors & limits

Common errors

Limits and notes

Related links