Create a chat completion

post

https://integrate.api.nvidia.com/v1/chat/completions

Supports text, up to eight images, function tools, and streaming. Video and audio inputs are unsupported. With stream=true, consume data-only SSE events until data: [DONE]; reasoning, content, and tool calls arrive as deltas.

Recent Requests

Time	Status	User Agent
Retrieving recent requests…

Loading…

Body Params

model

string

enum

required

Defaults to mistralai/mistral-large-4-1t-a52b

Allowed:

messages

array of objects

required

length ≥ 1

Conversation history. Use content-part arrays for images. After an assistant tool call, include a tool message with its tool_call_id. Preserve assistant reasoning_content when continuing a conversation.

Messages*

role

string

enum

required

Allowed:

content

string | null | array

reasoning_content

string | null

Assistant reasoning, separate from the final answer.

tool_calls

array of objects

Tool Calls

tool_call_id

string

Required on tool-result messages; matches an assistant tool call ID.

name

string

max_tokens

integer

1 to 262144

Defaults to 4096

Maximum generated tokens, including reasoning and the final answer. Small budgets may end during reasoning with null content and finish_reason=length. Prompt plus output must fit the context window.

stream

boolean

Defaults to false

temperature

number

0 to 2

Defaults to 0.5

The sampling temperature to use for text generation. The higher the temperature value is, the less deterministic the output text will be. It is not recommended to modify both temperature and top_p in the same call.

top_p

number

≤ 1

Defaults to 1

The top-p sampling mass used for text generation. The top-p value determines the probability mass that is sampled at sampling time. For example, if top_p = 0.2, only the most likely tokens (summing to 0.2 cumulative probability) will be sampled. It is not recommended to modify both temperature and top_p in the same call.

frequency_penalty

number

-2 to 2

Defaults to 0

Indicates how much to penalize new tokens based on their existing frequency in the text so far, decreasing model likelihood to repeat the same line verbatim.

presence_penalty

number

-2 to 2

Defaults to 0

Positive values penalize new tokens based on whether they appear in the text so far, increasing model likelihood to talk about new topics.

seed

integer

0 to 9007199254740991

Optional sampling seed, limited to JavaScript-safe integers for the interactive examples. Deterministic output is not guaranteed.

stop

string | array | null

Stop string or list of stop strings; omitted from the output.

tools

array of objects

Functions the model may request; the caller executes them.

Tools

tool_choice

string | object

Choose auto, none, required, or force a named function.

reasoning_effort

string

enum

Defaults to high

Controls reasoning for this request. Set to high, the model reasons first and the response carries reasoning_content separately from the final content; set to none, it answers directly. The playground sends high by default; a request that omits this field is served as none. chat_template_kwargs.enable_thinking is not honored by this recipe.

Allowed:

Headers

string

enum

Defaults to application/json

Generated from available response content types

Allowed:

Responses

400

Invalid request or unsupported input.

401

Missing or invalid NVIDIA API key.

402

Insufficient API credits.

422

Request validation failed.

429

Request rate limit exceeded.

Language

Credentials

Bearer

Loading…

Response

Click Try It! to start a request and see the response here! Or choose an example:

application/json

text/event-stream

200A chat completion, or SSE chunks when streaming is enabled.