Skip to navigation
BeatAPIMiniMaxText API

MiniMax Chat Completions

Call MiniMax M3 through one OpenAI Chat Completions endpoint with a BeatAPI key.

  • One endpoint, POST /v1/chat/completions, for MiniMax M3.
  • Streaming and function calling, with a 1M-token context window. Rates double once the prompt passes 512,000 tokens.
  • Works with the OpenAI SDK: change the base URL and use a BeatAPI key.
  • The same models also answer on OpenAI Responses and Anthropic Messages.

Use https://direct.beatapi.io as the base URL for text requests: it is not subject to Cloudflare’s 100-second limit, so a long non-streaming request is not cut off. https://api.beatapi.io also works and is the fallback if direct.beatapi.io is unreachable from your network. The key, models and prices are the same on both. Image and video APIs use https://api.beatapi.io only.

POST/v1/chat/completionsTry it

Opens this endpoint in the API Explorer. Enter your BeatAPI API key and send: the call is real and is billed to your own balance.

Models

Send this ID as model.

Model IDContext (tokens)Max output (tokens)PositioningInput, USD per 1M tokensOutput, USD per 1M tokensNotes
MiniMax-M31,048,576128,000MiniMax multimodal model for long-context coding, perception and agent planning$0.21$0.84MiniMax multimodal model for long-context coding, perception and agent planning. Rates double once the prompt passes 512,000 tokens.
  • Prices are the base tier, discount already applied. A long-context tier, cached-input rates and cache writes are not in this table; see the price table.
  • — means the catalogue does not publish the number.

Quick start

Install the openai package on your server, keep the key in BEATAPI_API_KEY, and never put it in browser code. These are request examples, not recorded model outputs.

Read the answer from choices[0].message.content, then check finish_reason and usage. An HTTP 200 alone does not prove the answer satisfies your task: validate it in your application. Text requests return directly; do not poll /v1/tasks for them.

curl --request POST \
  --url https://direct.beatapi.io/v1/chat/completions \
  --header "Authorization: Bearer $BEATAPI_API_KEY" \
  --header 'Content-Type: application/json' \
  --data '{
  "model": "MiniMax-M3",
  "max_tokens": 2000,
  "messages": [
    {
      "role": "user",
      "content": "Design a resilient webhook retry strategy for a payments API."
    }
  ]
}'

Endpoint and authentication

POST https://direct.beatapi.io/v1/chat/completions

Authorization string Required

Bearer <BEATAPI_API_KEY>. The key works with or without the sk- prefix.

Request parameters

Standard OpenAI Chat Completions fields are accepted. Request-format compatibility does not mean every model-specific option is supported, so test the options you rely on.

model string Required

The exact model ID.

Allowed values: MiniMax-M3

  • Each ID is a separate model, not an alias of another.
  • Call GET /v1/models with your key to confirm which models are enabled for it.
  • IDs are case-sensitive: send MiniMax-M3, not minimax-m3.

messages object[] Required

The conversation, oldest message first. Each message has a role (system, user, assistant or tool) and content.

max_tokens integer Optional

Upper bound on the generated tokens, reasoning included.

  • The gateway does not compare your value with the model’s max output from the table above. Keep it at or below that number.
  • max_tokens is also used to estimate how much balance a request needs. A very large value can be refused with 402 insufficient_credits even when the real output would be small. Set it to what you need.

A reasoning model writes its reasoning first. If max_tokens is too small, the reasoning can use all of it: content comes back empty and finish_reason is length. Leave room for both, and raise the value when you see length.

stream boolean OptionalDefault: false

Return the answer as server-sent events. See the Streaming section below.

Default: false

temperature number Optional

Sampling temperature. Change this or top_p, not both.

top_p number Optional

Nucleus sampling threshold.

stop string | string[] Optional

One or more sequences at which generation stops.

tools object[] Optional

Function definitions the model may call. The model answers with tool_calls; return each result in a following tool message that carries the matching tool_call_id.

tool_choice string | object Optional

Whether the model may, must or must not call a tool. Test the value you rely on with your model.

Long-context pricing

MiniMax-M3 moves to a higher rate once the prompt passes 512,000 tokens, and the higher rate applies to the whole request: input, cached input and output all double. It is a step, not a gradual increase: a request one token over the threshold bills entirely at the higher rate.

The prices in the model table above and on the price table are the base rates, for prompts under the threshold. Estimate the full prompt before you send a long history.

Response

A non-streaming call returns one JSON object. This is an abbreviated example.

{
  "id": "chatcmpl-example",
  "object": "chat.completion",
  "created": 1791993600,
  "model": "MiniMax-M3",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Retry with exponential backoff and jitter, and make every handler idempotent."
      },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 19,
    "completion_tokens": 16,
    "total_tokens": 35
  }
}

Response fields

FieldTypeDescription
idstringThe completion ID. Log it, together with the request ID, for support and billing audits.
modelstringThe model that answered.
choices[].message.contentstringThe answer text.
choices[].message.reasoning_contentstringThe reasoning text, on models that reason. It is billed as output.
choices[].message.tool_callsobject[]Function calls the model asks you to run. Present when finish_reason is tool_calls.
choices[].finish_reasonstringstop for a complete answer, tool_calls for a tool request, length when max_tokens was reached.
usageobjectToken counts. Calls are metered from actual token usage.

Streaming

Set "stream": true. The answer arrives as chat.completion.chunk events and ends with data: [DONE].

  • The response is text/event-stream; without streaming it is application/json.
  • Text arrives in choices[0].delta.content. A reasoning model sends its reasoning in delta.reasoning_content first.
  • A failure before the model is called is returned as a normal JSON error with an HTTP status, not as a stream event.
  • Set explicit timeouts and reconnect a streamed response safely.
data: {"id":"chatcmpl-example","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"role":"assistant","content":"Retry"},"finish_reason":null}]}
data: {"id":"chatcmpl-example","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":" with backoff"},"finish_reason":null}]}
data: {"id":"chatcmpl-example","object":"chat.completion.chunk","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}
data: [DONE]

Errors

Error or symptomCauseFix
content is empty and finish_reason is lengthA reasoning model used the whole max_tokens budget on reasoning.Raise max_tokens, within the model max output.
402 insufficient_creditsBalance below the price of the request, often because max_tokens is very large.Top up, or lower max_tokens.
403 forbiddenThe account is on free credit and the model is not on the free list, or the model is outside the key’s group.Top up, use a free model, or use a key whose group can reach the model.
404 not_foundUnknown model ID, or not enabled for this key.List the enabled models with GET /v1/models.
429 rate_limit_exceededAccount or free-model limit reached.Wait Retry-After, then retry.

HTTP errors at a glance

Authentication, routing, rate-limit and unknown-model failures are answered before the model is called, in the BeatAPI envelope: error.code, error.message, retryable, request_id. Every other failure keeps the OpenAI SDK shape {"error": {"message", "type", "code"}}: the code is in error.code and the request id ends message. See Errors.

HTTP statusCodeWhat to do
400bad_request, content_policy_violationFix the request or the input; do not resend it unchanged.
401missing_api_key, invalid_api_keyCheck the key. Do not retry.
402insufficient_creditsTop up, or raise the key’s limit. A very large max_tokens can trigger it.
403forbiddenThe model is outside the key’s group or IP allowlist, or free credit cannot call this model.
404not_foundUnknown model: list the enabled ones with GET /v1/models.
429rate_limit_exceededWait Retry-After or retry_after_seconds, then retry with backoff.
500internal_error, temporary_errorRetry once or twice with backoff.
503model_unavailable, upstream_timeout, temporary_errorRetry with backoff, or use another model.

Full code table and retry policy: Errors.

Notes

  • MiniMax also answers on OpenAI Responses and Anthropic Messages.
  • This page covers the MiniMax text model. The MiniMax H3 video models use the Video API.
  • Keep API keys on trusted servers, log the response request ID, and retry only 429, 500 and 503.

Authentication

AuthorizationBearer

Send your API key as Authorization: Bearer <BEATAPI_API_KEY>.

Request

This endpoint expects an object.
modelenumRequired

The exact model ID.

Each ID is a separate model, not an alias of another.

Call GET /v1/models with your key to confirm which models are enabled for it.

IDs are case-sensitive: send MiniMax-M3, not minimax-m3.

MiniMax-M3

Allowed values:
messageslist of objectsRequired

The conversation, oldest message first. Each message has a role (system, user, assistant or tool) and content.

max_tokensintegerOptional

Upper bound on the generated tokens, reasoning included.

The gateway does not compare your value with the model’s max output from the table above. Keep it at or below that number.

max_tokens is also used to estimate how much balance a request needs. A very large value can be refused with 402 insufficient_credits even when the real output would be small. Set it to what you need.

A reasoning model writes its reasoning first. If max_tokens is too small, the reasoning can use all of it: content comes back empty and finish_reason is length. Leave room for both, and raise the value when you see length.

streambooleanOptionalDefaults to false

Return the answer as server-sent events. See the Streaming section below.

temperaturedoubleOptional

Sampling temperature. Change this or top_p, not both.

top_pdoubleOptional
Nucleus sampling threshold.
stopstring or list of stringsOptional
One or more sequences at which generation stops.
toolslist of objectsOptional

Function definitions the model may call. The model answers with tool_calls; return each result in a following tool message that carries the matching tool_call_id.

tool_choicestring or map from strings to anyOptional
Whether the model may, must or must not call a tool. Test the value you rely on with your model.

Response

A non-streaming call returns one JSON object. This is an abbreviated example.

idstringOptional
The completion ID. Log it, together with the request ID, for support and billing audits.
modelstringOptional
The model that answered.
choiceslist of objectsOptional
usageobjectOptional
Token counts. Calls are metered from actual token usage.

Errors

400
Bad Request Error
401
Unauthorized Error
402
Payment Required Error
403
Forbidden Error
404
Not Found Error
429
Too Many Requests Error
500
Internal Server Error
503
Service Unavailable Error