Per-Model Details

How specific model families handle parameters differently from the defaults in Parameter Support by Model. All parameters below are top-level request fields unless noted — you can send them through any API format. The router normalises them automatically, so you generally don't need to know these details; they're here for when you want precise control.

GPT-5 series

GPT-5 models apply strict parameter sanitisation:

Parameter	Behaviour
`max_tokens`	Automatically converted to `max_completion_tokens`.
`temperature`	Accepted only when `= 1.0`, or when `reasoning_effort = "none"` (GPT-5.1+).
`presence_penalty`, `frequency_penalty`, `stop`, `logit_bias`	Always dropped.
`logprobs`, `top_logprobs`, `top_p`	Accepted only when `reasoning_effort = "none"` (GPT-5.1+).
`reasoning_effort`	Supported on all GPT-5 variants.

Claude 4+

The router normalises the thinking parameter to match the target Claude model — clients don't need to know which mode a model uses.

Adaptive thinking models (Claude Sonnet 4.6, Opus 4.6, Opus 4.7, Opus 4.8 and newer):

Parameter	Behaviour
`thinking.type = "enabled"`	Converted to `"adaptive"`. `budget_tokens` is optional and ignored when adaptive.
`thinking.type = "adaptive"`	Passed through directly.
`reasoning_effort`	Mapped to `thinking.type = "adaptive"`.

Non-adaptive thinking models (Claude Haiku 4.5, Sonnet 4.5, Opus 4.5):

Parameter	Behaviour
`thinking.type = "enabled"`	Requires `budget_tokens` (minimum 1024).
`thinking.type = "adaptive"`	Converted to `"enabled"` with a default budget.
`reasoning_effort`	Mapped to `budget_tokens` (low / medium / high → token counts).
`budget_tokens`	Must be ≥ 1024 and less than `max_tokens`.

All Claude models support tools, tool_choice, parallel_tool_calls, temperature, top_p, max_tokens, and response_format.

Gemini

Parameter	Behaviour
`thinking` / `reasoning_effort`	Accepted on all Gemini models; silently ignored on models without thinking support. Maps to a `thinking_level` tier on Gemini 3.x, to a token budget on Gemini 2.5 — see below.
`frequency_penalty`, `presence_penalty`	Not available on preview or Gemini 3 models.
`top_k`	Not a standard top-level field — pass via `extra_body`; forwarded when supported natively.

thinking_level support (Gemini 3.x only). Gemini 3.x models take a thinking_level tier instead of a token budget — sending an unsupported tier (e.g. medium on a model with only minimal/high) returns 400. Gemini 2.5 models don't use tiers at all; control them via reasoning.max_tokens / effort instead (see Reasoning & Thinking).

Model	Code	Supported `thinking_level` values	Default
Gemini 3.1 Flash-Lite	`gemini-3.1-flash-lite`	`minimal`, `low`, `medium`, `high`	`minimal`
Gemini 3 Flash Preview	`gemini-3-flash-preview`	`minimal`, `low`, `medium`, `high`	`high`
Gemini 3.1 Pro Preview	`gemini-3.1-pro-preview`	`low`, `medium`, `high` (no `minimal`)	`high`
Gemini 3.1 Flash Image	`gemini-3.1-flash-image`	`minimal`, `high` only	`minimal`
Gemini 3.1 Flash-Lite Image	`gemini-3.1-flash-lite-image`	`minimal`, `high` only	`minimal`
Gemini 3 Pro Image	`gemini-3-pro-image`	`high` only — any explicit value returns 400	`high`

Google's own docs also list plain Gemini 3 Pro and Gemini 3.1 Pro (low/medium/high, default high) as distinct model names — SIRAYA's catalog currently only exposes the preview build (gemini-3.1-pro-preview) on the text side. gemini-2.5-pro, gemini-2.5-flash, and gemini-2.5-flash-lite aren't in this table because the 2.5 series doesn't use thinking_level tiers at all (token-budget based, see above). gemini-2.5-flash-image isn't in this table because it doesn't support thinking at all (400). gemini-3.6-flash, gemini-3.5-flash, and gemini-3.5-flash-lite are not yet in this table — they're newer than Google's published thinking_level reference and their supported tiers/defaults haven't been confirmed yet. gemini-3.1-flash-lite-preview and gemini-3-pro-preview have been retired from SIRAYA's catalog and are omitted. See Nano Banana → extra_body for the image-model parameter details.