Support gemini-3.5-flash model in "google-vertex" provider

[sandesh-jagdale-tp]

What do you want to change?

Google has released the new "gemini-3.5-flash" model recently. I want to add support for that in "google-vertex" provider.

Why?

The addition of "gemini-3.5-flash" model in Google Vertex provider will allow pi users who are using Google Vertex as their provider, to use this model.

How? (optional)

No response

[github-actions]

This issue was auto-closed. All issues from new contributors are auto-closed by default.

Maintainers review auto-closed issues daily and reopen worthwhile ones. Issues that do not meet the quality bar in CONTRIBUTING.md will not be reopened or receive a reply.

If a maintainer replies lgtmi on one of your issues, your future issues will stay open. If a maintainer replies lgtm, your future issues and PRs will stay open.

See CONTRIBUTING.md.

[mitsuhiko]

Implemented the model catalog update for google-vertex.

Changes:

• google-vertex Gemini models now come from the models.dev Vertex catalog instead of a static list.
• Added gemini-3.5-flash.
• Kept gemini-flash-latest / gemini-flash-lite-latest and mapped their metadata to the current concrete models.
• Corrected Vertex gemini-2.5-flash cache-read pricing to match Google’s pricing page.
• Removed shut-down Vertex preview models, including gemini-3-pro-preview.
• Updated Google/Vertex provider thinking detection so latest aliases use Gemini 3 thinkingLevel behavior instead of the old budget path.

Validation:

• npm run check passes.

This comment is AI-generated by /wr

[skygrip]

Thanks for this fix! I believe this also fixes the previously incorrect hardcoded model pricing (beyond just the cache-read pricing).