### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

When using LiteLLM with Vertex Anthropic model `vertex_ai/claude-opus-4-7`, both parameters work independently, but fail when combined: - `thinking={"type":"adaptive"}` works alone - `response_format=<PydanticModel>` works alone - Both together fail with HTTP 400

## Repro matrix - thinking only: ✅ - response_format only: ✅ - thinking + response_format: ❌ (400)

## Additional context This appears to be in request translation for Vertex Anthropic, where `response_format` may map to forced `tool_choice`.

### Steps to Reproduce

from pydantic import BaseModel from litellm import completion

class DummyReport(BaseModel): msg: str

resp = completion( model="vertex_ai/claude-opus-4-7", messages=[{"role": "user", "content": "Say hello in Portuguese"}], vertex_credentials=<valid_vertex_credentials>, vertex_location="global", thinking={"type": "adaptive"}, response_format=DummyReport, ) print(resp)

### Relevant log output

```shell ## Error message `invalid_request_error: Thinking may not be enabled when tool_choice forces …