Chat Completions

Create text, multimodal, and tool-enabled responses with the OpenAI-compatible Chat Completions API.

Chat Completions generates a model response from a list of messages. Depending on the selected model, the endpoint can support text chat, multimodal input, streaming output, structured output, and function tools.

POST/v1/chat/completions

Generate a model response from the supplied messages. The endpoint supports standard and SSE streaming responses. Advanced capabilities such as image or audio input, tool calling, and structured output depend on the selected model.

When switching models, start with a minimal text request, then enable advanced parameters one at a time.

Authorization

bearerAuth

AuthorizationBearer <token>

Use Bearer your-api-key in the Authorization header.

In: header

Request Body

Choose a request structure based on the model input modality. All three views use the same Chat Completions endpoint and shared generation parameters.

application/json

Text chat

Use this mode for chat, content generation, and text reasoning. Message content can be a string.

model: A text model ID available to the current account.
messages[].role: The message role, such as user, system, or assistant.
messages[].content: The text message as a string.

application/json

{
  "model": "MODEL_ID",
  "messages": [
    {
      "role": "user",
      "content": "Hello! Introduce yourself in one sentence."
    }
  ]
}

Text chat request. User messages can be strings or content arrays containing only text.

model*string

The text model ID to use. Copy an ID visible to the current account that supports text input from the model marketplace.

Length1 <= length

messages*array<|||||>

Text conversation messages in the order they are sent.

Items1 <= items

store?boolean|null

Whether to store the output of this chat completion. Availability depends on the selected model and provider channel.

Defaultfalse

reasoning_effort?string|null

Reasoning effort for reasoning models. Available values depend on the selected model; none explicitly disables reasoning effort.

Default"medium"

Value in

"none"
"minimal"
"low"
"medium"
"high"
"xhigh"
"max"
null

metadata?|null

Key-value metadata attached to the request. Keys can contain up to 64 characters and values up to 512 characters.

Propertiesproperties <= 16

modalities?array<>|null

Output types the model should generate. By default, the model generates text only.

prediction?|null

Predicted content configuration for cases where most of the expected output is already known.

audio?|null

Audio format and voice configuration used when requesting audio output.

temperature?number|null

Sampling temperature. Usually, adjust either temperature or top_p, but not both. Support depends on the selected model.

Range0 <= value <= 2

Default1

top_p?number|null

Nucleus sampling probability mass. Usually, adjust either temperature or top_p, but not both. Support depends on the selected model.

Range0 <= value <= 1

Default1

n?integer|null

Number of response choices to generate for each input. Generating multiple choices increases usage.

Range1 <= value <= 128

Default1

stop?string|array<string>|null

Up to four stop sequences. Support and exact behavior depend on the selected model.

max_tokens?integer|null

Deprecated

Legacy maximum output token parameter. Use max_completion_tokens instead.

Range1 <= value

max_completion_tokens?integer|null

Maximum number of tokens the completion may generate, including visible output and reasoning tokens.

presence_penalty?number|null

Penalizes tokens based on whether they have already appeared, increasing the likelihood of introducing new topics. Support depends on the selected model.

Range-2 <= value <= 2

Default0

frequency_penalty?number|null

Penalizes tokens based on how often they have appeared, reducing the likelihood of repeated content. Support depends on the selected model.

Range-2 <= value <= 2

Default0

logit_bias?|null

Maps token IDs to bias values from -100 to 100.

logprobs?boolean|null

Whether to return log probabilities for output tokens. Support depends on the selected model.

Defaultfalse

user?string

Deprecated

Deprecated end-user identifier. Use safety_identifier for safety monitoring and prompt_cache_key for cache routing.

safety_identifier?string|null

Stable end-user identifier for safety and abuse monitoring. Use a hashed value that does not contain personal information.

Lengthlength <= 64

prompt_cache_key?string|null

Stable key used to improve cache hits for similar requests. Do not include personal or other sensitive data.

prompt_cache_options?|null

Request-level prompt caching configuration.

prompt_cache_retention?string|null

Deprecated

Deprecated cache retention setting. Use prompt_cache_options.ttl instead.

Value in

"in_memory"
"24h"
null

verbosity?string|null

Controls the level of detail in text responses. Support depends on the selected model.

Default"medium"

Value in

"low"
"medium"
"high"
null

moderation?|null

Configures content moderation policies for input and output.

service_tier?string|null

Service tier used for the request. Availability depends on the account and upstream capabilities.

Default"auto"

Value in

"auto"
"default"
"flex"
"scale"
"priority"
"fast"
null

stream_options?|null

Streaming response options used only when stream: true.

response_format?||

Specifies the model output format. Structured output support depends on the selected model.

seed?integer|null

Deprecated

Beta parameter that makes a best effort at deterministic sampling, without guaranteeing deterministic output.

tools?array<|>

Tools the model may call.

Itemsitems <= 128

functions?array<>

Deprecated

Legacy function definitions. Use tools instead.

tool_choice?|||

Controls whether the model calls a tool, requires a specific tool, or limits the available tools to an allowed set.

function_call?|

Deprecated

Legacy function selection parameter. Use tool_choice instead.

parallel_tool_calls?boolean

Whether the model may call multiple tools in parallel within one response.

Defaulttrue

stream?boolean|null

When true, returns chat completion chunks incrementally through Server-Sent Events.

Defaultfalse

top_logprobs?integer|null

Number of most likely tokens returned at each token position. Set logprobs: true when using this field. Support depends on the selected model.

Range0 <= value <= 20

web_search_options?

Image and audio inputs are available only for models that support the corresponding modality. Check the models available to the current account and the actual API capabilities.

Response Body

`application/json`

curl -X POST "https://example.com/v1/chat/completions" \  -H "Content-Type: application/json" \  -d '{    "model": "MODEL_ID",    "messages": [      {        "role": "user",        "content": "Hello! Introduce yourself in one sentence."      }    ]  }'

{
  "id": "chatcmpl_example",
  "object": "chat.completion",
  "created": 1741569952,
  "model": "MODEL_ID",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "Hello! How can I help you?",
        "refusal": null,
        "annotations": []
      },
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 19,
    "completion_tokens": 10,
    "total_tokens": 29,
    "prompt_tokens_details": {
      "cached_tokens": 0,
      "audio_tokens": 0
    },
    "completion_tokens_details": {
      "reasoning_tokens": 0,
      "audio_tokens": 0,
      "accepted_prediction_tokens": 0,
      "rejected_prediction_tokens": 0
    }
  }
}

OpenAI Chat Completions documentationReview the upstream protocol and recent field changes.

Authorization

Request Body

Response Body

200application/json

400application/json

401application/json

403application/json

404application/json

429application/json

500application/json

502application/json

503application/json

504application/json

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`

`application/json`