Unique AI Documentation

GPT-5.6 Usage with the Responses API

2 min read

How to use GPT-5.6 model with UniqueAI

Models deployed via Azure OpenAI

For GPT-5.6 models deployed via Azure OpenAI, Unique enables the Responses API by default. No additional configuration is required.

Models deployed directly via OpenAI through LiteLLM

For GPT-5.6 models deployed directly via OpenAI through LiteLLM, enable the Responses API manually in the model's advanced configuration.

Where to configure it

  1. Open the space configuration in Unique.

  2. Go to Advanced Configuration.

  3. Go to Loop Agent.

  4. Go to Experimental.

  5. Enable Use Responses API.

  6. Save the configuration.

If this option is not enabled, GPT-5.6 requests may not work correctly.

Setting the Reasoning Effort

Reasoning effort is configured in the space configuration under Advanced Configuration > Loop Agent > Experimental > Additional LLM Options.

Both spellings are accepted — regardless of whether the Responses API is enabled:

Key

Example value

reasoning_effort

high

reasoning

{"effort": "high"}

The platform translates the flat reasoning_effort key into the Responses reasoning.effort parameter before the request is sent and removes the flat key from the payload. You therefore do not need to rename the key when the Responses API is active — existing configurations keep working.

Precedence: configure only one of the two keys. If both reasoning and reasoning_effort are set, the reasoning object takes precedence and reasoning_effort is silently ignored.

Validation: if the configured value is not supported by the selected model, the request does not fail. The platform logs a warning and falls back to the lightest reasoning effort the model supports.

Temperature: while an active reasoning effort (any value other than none) is set, the platform automatically forces the temperature to 1, as required by the API. A configured temperature has no effect in that case.

Default: GPT-5.6 (Sol, Terra, Luna) defaults to none — the model does not reason at all unless an effort is explicitly configured in the space. Supported values are none, low, medium, high, and xhigh.

Language Model Info

If the model is used before release 2026.30, the complete language model info has to be set. Only using the enum will use default parameters, that deviate significantly from the one of GPT 5.6 models

pGPT 5.6 Sol

py
{
    "name": "litellm:openai-gpt-5-6-sol",
    "provider": "LITELLM",
    "version": "2026-07-09",
    "capabilities": [
        "chat_completions_api",
        "function_calling",
        "parallel_function_calling",
        "reasoning",
        "responses_api",
        "streaming",
        "structured_output",
        "vision",
    ],
    "token_limits": {
        "token_limit_input": 922_000,
        "token_limit_output": 128_000,
    },
    "default_options": {"reasoning_effort": "none"},
    "supported_reasoning_efforts": [
        "none",
        "low",
        "medium",
        "high",
        "xhigh",
    ],
}

GPT 5.6 Terra

py
{
    "name": "litellm:openai-gpt-5-6-terra",
    "provider": "LITELLM",
    "version": "2026-07-09",
    "capabilities": [
        "chat_completions_api",
        "function_calling",
        "parallel_function_calling",
        "reasoning",
        "responses_api",
        "streaming",
        "structured_output",
        "vision",
    ],
    "token_limits": {
        "token_limit_input": 922_000,
        "token_limit_output": 128_000,
    },
    "default_options": {"reasoning_effort": "none"},
    "supported_reasoning_efforts": [
        "none",
        "low",
        "medium",
        "high",
        "xhigh",
    ],
}

GPT 5.6 Luna

py
{
    "name": "litellm:openai-gpt-5-6-luna",
    "provider": "LITELLM",
    "version": "2026-07-09",
    "capabilities": [
        "chat_completions_api",
        "function_calling",
        "parallel_function_calling",
        "reasoning",
        "responses_api",
        "streaming",
        "structured_output",
        "vision",
    ],
    "token_limits": {
        "token_limit_input": 922_000,
        "token_limit_output": 128_000,
    },
    "default_options": {"reasoning_effort": "none"},
    "supported_reasoning_efforts": [
        "none",
        "low",
        "medium",
        "high",
        "xhigh",
    ],
}
Last updated