VLLM

Create chat completion (OpenAI-compatible)

post

Create a chat completion using vLLM (OpenAI-compatible API).

Available Models: openai/gpt-oss-120b

Required Parameters

Parameter	Type	Description
model	string	Model ID: openai/gpt-oss-120b
messages	array	Array of message objects with role and content

Optional Parameters

Parameter	Type	Default	Description
reasoning_effort	string	-	Reasoning depth: low, medium, high. Omit to use model default.
include_reasoning	boolean	-	Include reasoning in response. Omit to use model default.
max_completion_tokens	integer	-	Max tokens to generate (preferred)
max_tokens	integer	-	Max tokens (deprecated, use max_completion_tokens)
min_tokens	integer	-	Minimum tokens before stopping
temperature	float	-	Sampling temperature (0.0-2.0)
top_p	float	-	Nucleus sampling threshold (0.0-1.0)
top_k	integer	-	Top-k sampling (-1 to disable)
min_p	float	-	Minimum probability threshold (0.0-1.0)
frequency_penalty	float	-	Penalize frequent tokens (-2.0 to 2.0)
presence_penalty	float	-	Penalize repeated topics (-2.0 to 2.0)
repetition_penalty	float	-	Repetition penalty (1.0 = none)
n	integer	-	Number of completions (1-10)
stop	string/array	-	Stop sequence(s)
seed	integer	-	Random seed for reproducibility
logprobs	boolean	-	Return log probabilities
top_logprobs	integer	-	Number of top logprobs (0-20)
logit_bias	object	-	Token bias mapping
user	string	-	End-user identifier

Async Mode

Set async: true to get a task_address immediately and poll for results.

Header parameters

x-api-keyany ofOptional

stringOptional

or

nullOptional

Body

application/json

OpenAI-compatible chat completion request for vLLM.

Supports standard OpenAI API parameters plus vLLM-specific extensions. Parameters are passed directly to the vLLM provider without transformation.

modelstringRequired

Model identifier. Available: openai/gpt-oss-120b

Example: openai/gpt-oss-120b

max_tokensany ofOptional

Maximum tokens to generate (deprecated, use max_completion_tokens)

integer · min: 1 · max: 32768Optional

or

nullOptional

max_completion_tokensany ofOptional

Maximum tokens to generate (preferred over max_tokens)

integer · min: 1 · max: 32768Optional

or

nullOptional

min_tokensany ofOptional

Minimum tokens to generate before stopping

integerOptional

or

nullOptional

temperatureany ofOptional

Sampling temperature (0.0 = deterministic)

number · max: 2Optional

or

nullOptional

top_pany ofOptional

Top-p (nucleus) sampling

number · max: 1Optional

or

nullOptional

top_kany ofOptional

Top-k sampling (-1 to disable)

integer · min: -1Optional

or

nullOptional

min_pany ofOptional

Minimum probability threshold for sampling

number · max: 1Optional

or

nullOptional

frequency_penaltyany ofOptional

Frequency penalty for token repetition

number · min: -2 · max: 2Optional

or

nullOptional

presence_penaltyany ofOptional

Presence penalty for topic repetition

number · min: -2 · max: 2Optional

or

nullOptional

repetition_penaltyany ofOptional

Repetition penalty (1.0 = no penalty)

numberOptional

or

nullOptional

nany ofOptional

Number of completions to generate

integer · min: 1 · max: 10Optional

or

nullOptional

stopany ofOptional

Stop sequence(s) - generation stops when encountered

stringOptional

or

string[]Optional

or

nullOptional

seedany ofOptional

Random seed for reproducibility

integerOptional

or

nullOptional

streamany ofOptional

Enable streaming responses (not yet supported)

Default: false

booleanOptional

or

nullOptional

reasoning_effortany ofOptional

Reasoning effort level: 'low', 'medium', 'high'. Omit to use model default.

string · enumOptionalPossible values:

or

nullOptional

include_reasoningany ofOptional

Include reasoning content in response. Omit to use model default.

booleanOptional

or

nullOptional

logprobsany ofOptional

Return log probabilities of output tokens

booleanOptional

or

nullOptional

top_logprobsany ofOptional

Number of most likely tokens to return at each position

integer · max: 20Optional

or

nullOptional

logit_biasany ofOptional

Token ID to bias value mapping (-100 to 100)

or

nullOptional

userany ofOptional

Unique identifier for the end-user

stringOptional

or

nullOptional

modeany ofOptional

Routing mode: 'auto' or 'direct' (vLLM is direct-only)

Default: auto

string · enumOptionalPossible values:

or

nullOptional

asyncany ofOptional

Async mode: returns task_address immediately, poll /v2/tasks/{task_address} for result. Default: false (sync mode).

Default: false

booleanOptional

or

nullOptional

Responses

200

Successful Response

application/json

202

Task accepted (async mode). Poll the poll_url for status.

application/json

422

Validation Error

application/json

post

/v2/vllm/v1/chat/completions

HTTPcURLJavaScriptPython

POST /v2/vllm/v1/chat/completions HTTP/1.1
Content-Type: application/json
Accept: */*
Content-Length: 92

{
  "model": "openai/gpt-oss-120b",
  "messages": [
    {
      "role": "user",
      "content": "Hello, how are you?"
    }
  ]
}

Minimal (Recommended)With Reasoning ControlWith Sampling ParamsAll Options

{
  "id": "text",
  "object": "chat.completion",
  "created": 1,
  "model": "text",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "text",
        "reasoning_content": "text",
        "refusal": "text",
        "ANY_ADDITIONAL_PROPERTY": "anything"
      },
      "finish_reason": "text",
      "logprobs": null,
      "ANY_ADDITIONAL_PROPERTY": "anything"
    }
  ],
  "usage": {
    "prompt_tokens": 0,
    "completion_tokens": 0,
    "total_tokens": 0,
    "ANY_ADDITIONAL_PROPERTY": "anything"
  },
  "system_fingerprint": "text",
  "task_address": "text",
  "mode": "text",
  "ANY_ADDITIONAL_PROPERTY": "anything"
}

List available models (OpenAI-compatible)

get

List all available vLLM models (OpenAI-compatible format).

Returns all active vLLM models. Access control is enforced at request time via tier model_restrictions - this endpoint shows what's available.

Responses

200

Successful Response

application/json

get

/v2/vllm/v1/models

HTTPcURLJavaScriptPython

GET /v2/vllm/v1/models HTTP/1.1
Accept: */*

200

Successful Response

{
  "object": "list",
  "data": [
    {
      "id": "openai/gpt-oss-120b",
      "object": "model",
      "created": 1,
      "owned_by": "opengpu",
      "ANY_ADDITIONAL_PROPERTY": "anything"
    }
  ],
  "ANY_ADDITIONAL_PROPERTY": "anything"
}