PSA: Gemini 2.5 Pro and 2.5 Flash retire on December 1

Image: Google

If your code still calls gemini-2.5-pro or gemini-2.5-flash, change the model string before December 1. Google says requests to both models will be automatically redirected after that date, and billed at the new model's rates.

The company announced the deprecation in its AI Studio release notes. gemini-2.5-pro will route to gemini-3.8-flash, and gemini-2.5-flash will route to gemini-3.5-flash-lite. Neither 2.5 model is available to select in AI Studio anymore.

The Pro redirect lands on a Flash model, not another Pro model. Test the prompts and tool calls that matter before the switch, especially if you picked 2.5 Pro for long-form writing or complex reasoning.

Google lists these breaking API changes:

• Function calling and thought_signature validation change.
• thinking_level replaces thinking_budget.
• The sampling parameters temperature, top_p and top_k are removed.
• Prefilled model turns are removed.

The changes apply to both migration paths, though the details differ by model. Google has separate migration guides for 3.8 Flash and 3.5 Flash-Lite.

PSA: Upcoming Gemini API deprecations - Gemini 2.5 Pro will be deprecated and auto-redirected to 3.8 Flash on Dec 1 - Gemini 2.5 Flash will be deprecated and auto-redirected to 3.5 Flash-Lite on Dec 1

— Patrick Loeber (@patloeber) November 3, 2026

Check four things in your integration: the model string, any function or tool schemas, the thinking config, and any code that prefills an assistant turn. The redirect keeps old calls working after December 1, but it does not make them compatible.

The safest move is to switch now and compare outputs on a small test set. A silent model change is harder to debug after the fact.