Gemini 2.5 Pro and Flash migration: what changes December 1
Google’s November 2, 2026 announcement separates two changes developers should not confuse: Gemini 2.5 Pro and Flash are already unavailable to select in Google AI Studio, but automatic API rerouting is scheduled for December 1. Official changelog
Which model goes where?
| Existing API identifier | Scheduled destination from December 1, 2026 |
|---|---|
gemini-2.5-pro | gemini-3.8-flash |
gemini-2.5-flash | gemini-3.5-flash-lite |
Google says requests will use the destination model’s rates. Existing access to these older API models is restricted to past active users. Crucially, gemini-2.5-flash-lite is not deprecated and remains available through the API until further notice. This notice concerns the two identifiers above, not every model carrying the 2.5 name. Deprecation details
Budget for the destination, not just the alias
These are standard paid-tier USD prices per million tokens. Output includes thinking tokens. They are not Batch or Priority rates. Official pricing
| Destination | Input | Output |
|---|---|---|
| Gemini 3.8 Flash, through December 31, 2026 | $0.75 | $3.75 |
| Gemini 3.8 Flash, from January 1, 2027 | $1.50 | $7.50 |
| Gemini 3.5 Flash-Lite, current listed rates | $0.30 | $2.50 |
For comparison, 2.5 Pro currently costs $1.25 input and $10 output for prompts up to 200,000 tokens, rising to $2.50 and $15 above that. 2.5 Flash already lists $0.30 text input and $2.50 output, but audio input is $1 rather than the destination’s $0.30.
The Flash-Lite input rate covers text, image, video and audio. Tool use and caching can add separate charges. Unit prices alone do not establish the total cost of a workload.
Check requests before the cutover
Google’s migration guidance calls for replacing thinking_budget with thinking_level, removing unsupported request settings and preserving thought signatures during tool use. GenerateContent function responses need both call_id and name.
Gemini 3.8 Flash supports low, medium and high thinking. Medium is the default and minimal is unsupported. Gemini 3.5 Flash-Lite defaults to minimal. Google’s guide recommends medium or high for tool-heavy reasoning.
Do not treat every compatibility issue as the same error. The Flash-Lite migration guide says temperature, top_p and top_k are currently ignored, with errors planned for future generations. A prefilled model turn already returns HTTP 400. candidate_count is unsupported in Gemini 3+.
Before changing production, review prompts, structured output and multi-turn tool flows against the target. That is an integration checklist, not a claim that ByteForward has benchmarked either replacement. 3.8 migration guide · 3.5 Flash-Lite migration guide
AI-generated conceptual illustration by ByteForward, not a product interface.
