Thinking level support for Gemini 3 Pro

[maciej-nexos]

Description: (required)
Gemini 3 Pro model supports thinking_level parameter.
This new parameter is mutually exclusive with existing thinking_budget parameter for Gemini 2.5 series models.

Tests Run/Test cases added: (required)

• n/a

Type of Change:

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional changes)

[roh26it]

Thanks for this, @maciej-nexos

Will review this quickly and merge

[Copilot]

Pull Request Overview

This PR adds support for the thinking_level parameter in Gemini 3 Pro model, which is mutually exclusive with the existing thinking_budget parameter used by Gemini 2.5 series models.

• Added thinking_level field to the Params interface
• Updated Google and Google Vertex AI providers to handle thinking_level with mutual exclusivity logic
• Modified include_thoughts condition to accept either budget_tokens or thinking_level

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
src/types/requestBody.ts	Added optional thinking_level field to the thinking configuration type
src/providers/google/chatComplete.ts	Updated transformation logic to prioritize thinking_level over budget_tokens when present
src/providers/google-vertex-ai/transformGenerationConfig.ts	Applied same mutual exclusivity logic for Vertex AI provider

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

When thinking_level is not provided but budget_tokens is undefined or 0, this will set thinking_budget to a falsy value. This could cause issues if the API expects either a valid budget or no field at all. Consider only setting thinking_budget when budget_tokens has a meaningful value.

Suggested change

	thinkingConfig['thinking_budget'] = budget_tokens;
	}
	if (budget_tokens != null && budget_tokens !== 0) {
	thinkingConfig['thinking_budget'] = budget_tokens;
	}

[Copilot]

When thinking_level is not provided but budget_tokens is undefined or 0, this will set thinking_budget to a falsy value. This could cause issues if the API expects either a valid budget or no field at all. Consider only setting thinking_budget when budget_tokens has a meaningful value.

Suggested change

	thinkingConfig['thinking_budget'] = budget_tokens;
	}
	if (budget_tokens != null && budget_tokens !== 0) {
	thinkingConfig['thinking_budget'] = budget_tokens;
	}

[narengogi]

I think it would be best to map reasoning_effort to thinking_level

[narengogi]

I've updated the PR to map reasoning_effort to thinkingLevel

@b4s36t4 for review

[davidwhite-spec]

We would really love to see this feature added as users of Portkey.

[maciej-nexos]

anything blocking this from being merged in?

[narengogi]

merging this @davidwhite-spec and @maciej-nexos