Gemini 2.5 Pro calls will be answered by Gemini 3.8 Flash from December 1, and 2.5 Flash by 3.5 Flash-Lite

Google will start answering gemini-2.5-pro and gemini-2.5-flash API calls with different models on December 1, 2026. According to the Gemini API changelog entry dated November 2 and the deprecations page, requests to gemini-2.5-pro will be routed automatically to gemini-3.8-flash and billed at Gemini 3.8 Flash rates, and requests to gemini-2.5-flash will go to gemini-3.5-flash-lite. Neither model string stops working; it just stops meaning what it says.

Both 2.5 models are already unselectable in AI Studio, and API access is limited to existing users. gemini-2.5-flash-lite is not on the list.

Not a like-for-like swap

For 2.5 Pro users the new target is a Flash model, and for 2.5 Flash users it is a Flash-Lite model, so in both cases the replacement sits a tier down in Google's own naming. Whether that is a downgrade in practice depends on the workload; Gemini 3.8 Flash is a much newer model than 2.5 Pro. But nobody choosing a model by name would read "Pro" and expect "Flash".

Prices per million tokens, from Google's pricing page:

So the rerouted Pro traffic gets cheaper for one month, then the input price goes above what 2.5 Pro charged (output stays lower). Token counts may move too: Google's 3.8 documentation says the model "can use more tokens on longer running and complex tasks, by design", and its default thinking level is medium. For 2.5 Flash the list price is unchanged.

What breaks

The 3.8 migration guide lists changes that matter for code written against 2.5:

• thinking_budget is replaced by thinking_level (low, medium or high; minimal is not supported).
• temperature, top_p, top_k and candidate_count are removed, and so are prefilled model turns.
• Function calling now validates thought_signature, and FunctionResponse objects need call_id and name.

What the docs do not say is what happens to a legacy 2.5 Pro request that still sends temperature or thinking_budget once it lands on 3.8 Flash: whether it errors or the parameter is silently dropped. Anyone running 2.5 Pro in production with sampling settings or function calling should test against gemini-3.8-flash directly before December 1 rather than find out from the logs.

A pattern, not a one-off

This is the second time in less than a month that Google has handled a deprecation by aliasing. On October 8 it routed gemini-3.7-flash to gemini-3.8-flash and gemini-3.5-flash to gemini-3.6-flash. Retiring an old model by quietly pointing its name at a new one keeps applications from crashing on the cutoff date, which is the upside. The downside is that a pinned model version no longer guarantees the model, the price or the accepted parameters, and the only notice is a changelog line.