OpenAI-compat API rejects reasoning_effort="minimal" for Gemini 3 Flash

Gemini API
, , ,
1

Description:
When calling the OpenAI compatibility endpoint POST https://generativelanguage.googleapis.com/v1beta/openai/chat/completions with model gemini-3-flash* and reasoning_effort: "minimal", the API returns:

400 INVALID_ARGUMENT: Invalid reasoning_effort: minimal. Valid values are: high, low, medium, none

However, your OpenAI-compat documentation’s “Thinking” mapping table indicates reasoning_effort=minimal is supported/mapped for Gemini 3 Flash.

Please clarify if my understanding is correct, and if I should respect the API error response, or the documentation.

2 Likes
2

I am also facing the similar problem using langchain to call gemini-3-flash-preview with thinking_level=”minimal”.

1 Like