Google deprecates Gemini 2.5 Pro, routing API traffic to Gemini 3.8 Flash on December 1
Google says gemini-2.5-pro is deprecated and will automatically route to gemini-3.8-flash, billed at Gemini 3.8 Flash rates, starting December 1, 2026.
From December 1, 2026, Google will send all API requests for gemini-2.5-pro to gemini-3.8-flash, and those requests will be billed at Gemini 3.8 Flash rates. Google's Gemini API changelog, dated November 2, 2026, says gemini-2.5-pro is deprecated and is already no longer available to select in Google AI Studio. The redirect is automatic. Developers who do nothing will have their calls answered by a different model. The changelog asks developers to change their model string to gemini-3.8-flash before December 1 and to review the Gemini 3.8 Flash migration guide.
The changelog says the migration involves breaking API changes. It names three: function calling and thought_signature validation, thinking_level replacing thinking_budget, and removed sampling parameters and prefilled model turns.
The same changelog also deprecates gemini-2.5-flash, which will route to gemini-3.5-flash-lite on the same date under the same breaking-change list. That second deprecation is separate from the 2.5 Pro one. The changelog gives it its own migration target and pricing tier.
What changed
gemini-2.5-pro is deprecated, and its API requests are routed to gemini-3.8-flash and billed at Gemini 3.8 Flash rates.
Type: Breaking change · Takes effect: Tuesday, December 1, 2026 · Applies to: gemini-2.5-pro, gemini-3.8-flash
Why it matters
Teams still calling gemini-2.5-pro will see different behavior and pricing after December 1 unless they migrate. The listed breaking changes can affect tool use and parameter handling.
What you can do
Switch the model string to gemini-3.8-flash before December 1, 2026, and review the migration guide for the listed breaking changes.
How the story developed
The official changelog adds the list of breaking API changes, including function calling, thought_signature validation, thinking_level replacing thinking_budget, and removed sampling parameters and prefilled model turns.
Original source
Where we saw it
Gemini API changelog
AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works