App Version

v3.14.3

API Provider

GCP Vertex AI

Model Used

gemini-2.5-flash-preview-04-17:thinking

Actual vs. Expected Behavior

Actual: Max Thinking Tokens slider is shown for gemini-2.5-flash-preview-04-17:thinking with no option to not provide a value.

Expected: Provide checkbox (default on) to let gemini-2.5-flash determine its own thinking budget; this may frequently be the most performant mode of operation (cf https://simonwillison.net/2025/Apr/17/start-building-with-gemini-25-flash/)

If it is easier to execute, simply remove the Max Thinking Tokens slider for this model, swap backend to always allow the model to choose its own thinking budget, and consider adding checkbox at later date iff desired. Current design arbitrarily picks a non-default config that is likely net-harmful to output quality.

Detailed Steps to Reproduce

N/A

Relevant API Request Output

N/A

Additional Context

N/A