Google formally deprecated Gemini 2.5 Pro and Gemini 2.5 Flash on November 2, 2026. According to the Gemini API changelog, neither model can now be selected in Google AI Studio, and from December 1 API requests sent to the old model names will automatically be routed to new models and billed at the new models' rates. For developers still using the 2.5 series, this is not an ordinary version notice but a breaking change that must be migrated before December.
After rerouting, both the destination and the bill change
The changelog sets out two different rerouting paths. gemini-2.5-pro will automatically route to gemini-3.8-flash and be billed at Gemini 3.8 Flash rates, while gemini-2.5-flash will route to gemini-3.5-flash-lite and be billed at that model's rates. Google asks developers to change the model string to the new name before December 1 and to check against each migration guide. Automatic routing keeps requests flowing; it does not guarantee identical behavior, and the billing basis changes too. The Pro-to-Flash move deserves particular attention: the model's positioning has changed, so output quality, speed, and cost need to be measured again rather than assumed from earlier evaluations.
Migration means changing more than the model name
Both models share the same list of breaking changes, covering function calling and thought_signature validation, thinking_level replacing thinking_budget, and the removal of some sampling parameters and prefilled model turns. Simply swapping the model string and shipping is likely to fail outright on function calling or thinking configuration, or to change model behavior silently. The safer order is to run the three categories of cases covering function calling, thinking level, and sampling configuration in a test environment first, confirm that output and cost meet expectations, and only then switch production traffic.
Who should act first, before December
Teams that hard-code model names in code constants or configuration files, long-running services, and teams relying on 2.5 Pro's capabilities should go first. Individual users who pick models manually in AI Studio should also check whether the default selection has changed. After December 1, an old name still working does not mean the old model is still working: billing and behavior follow the new model, and discovering a problem on the day of the switch will only cost more.
