Struct ParameterSupport 

pub struct ParameterSupport {
    pub temperature: ParameterRestriction,
    pub frequency_penalty: ParameterRestriction,
    pub presence_penalty: ParameterRestriction,
    pub logprobs: bool,
    pub top_logprobs: bool,
    pub logit_bias: bool,
    pub n_multiple: bool,
    pub top_p: ParameterRestriction,
    pub reasoning: bool,
}

Parameter support information for a model.

This struct provides detailed information about which parameters are supported by a model and any restrictions that apply.

Example

use openai_tools::common::models::{ChatModel, ParameterRestriction};

let model = ChatModel::O3Mini;
let support = model.parameter_support();

// Reasoning models only support temperature = 1.0
assert_eq!(support.temperature, ParameterRestriction::FixedValue(1.0));

// Reasoning models don't support logprobs
assert!(!support.logprobs);

Fields

temperature: ParameterRestriction

Temperature parameter restriction (Chat & Responses API)

frequency_penalty: ParameterRestriction

Frequency penalty parameter restriction (Chat API only)

presence_penalty: ParameterRestriction

Presence penalty parameter restriction (Chat API only)

logprobs: bool

Whether logprobs parameter is supported (Chat API only)

top_logprobs: bool

Whether top_logprobs parameter is supported (Chat & Responses API)

logit_bias: bool

Whether logit_bias parameter is supported (Chat API only)

n_multiple: bool

Whether n > 1 (multiple completions) is supported (Chat API only)

top_p: ParameterRestriction

Top P parameter restriction (Responses API only)

reasoning: bool

Whether reasoning parameter is supported (Responses API only, reasoning models)

Implementations

impl ParameterSupport

pub fn standard_model() -> Self

Creates parameter support info for standard (non-reasoning) models.

Standard models support all parameters with full range.

pub fn reasoning_model() -> Self

Creates parameter support info for reasoning models (GPT-6, GPT-5, o-series).

Reasoning models have restricted parameter support:

• temperature: only 1.0
• top_p: only 1.0
• frequency_penalty: only 0
• presence_penalty: only 0
• logprobs, top_logprobs, logit_bias: not supported
• n: any value (verified against the API in September 2026 for the o-series and GPT-5 through GPT-6, all of which return n choices)
• reasoning: supported

pub fn search_model() -> Self

Creates parameter support info for web-search models (gpt-5-search-api, gpt-4o-search-preview, …).

These reject the whole sampling parameter set the same way reasoning models do, and unlike reasoning models they also reject n > 1. They expose no reasoning parameter - the API rejects both with “Model incompatible request argument supplied”.

Trait Implementations

impl Clone for ParameterSupport

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ParameterSupport

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.