On Nov. 2, 2026, Google added two deprecation notices to the Gemini API release notes: stable gemini-2.5-pro and gemini-2.5-flash are no longer available to pick in Google AI Studio, and starting Dec. 1, 2026, the API will silently reroute requests that still use those model strings to newer endpoints—and bill them at those models' published rates. Google is urging developers to update model IDs and read the migration guides before the cutoff.
Key facts
| Item | Detail |
|---|---|
| Announced | Nov. 2, 2026 |
| Effective | Dec. 1, 2026 |
gemini-2.5-pro | Routes to gemini-3.8-flash |
gemini-2.5-flash | Routes to gemini-3.5-flash-lite |
| Not affected | gemini-2.5-flash-lite |
What Google changed
Both models are marked deprecated on the release notes page dated Nov. 2, 2026. Google states that each is “no longer available to select in Google AI Studio.” The deprecations page lists June 17, 2025 as the release date for gemini-2.5-pro and gemini-2.5-flash, with a shutdown date of Dec. 1, 2026—the day automatic routing and rebilling kick in for API traffic that still names the old IDs.
Where each model routes and how it is billed
For gemini-2.5-pro, Google says that effective Dec. 1, 2026, all API requests will automatically route to gemini-3.8-flash and be billed at Gemini 3.8 Flash rates. Developers should update the model string to gemini-3.8-flash before that date and review the Gemini 3.8 Flash migration guide linked from the release notes.
For gemini-2.5-flash, the same Dec. 1 cutoff applies, but the destination is gemini-3.5-flash-lite, billed at Gemini 3.5 Flash-Lite rates. Google asks teams to switch the model string to gemini-3.5-flash-lite and read the Gemini 3.5 Flash-Lite migration guide before Dec. 1.
The pages link to Google's pricing tables for each target model; they do not compare old and new costs. If you rely on either 2.5 ID, treat billing as moving to the replacement model's rate card and verify spend against your own usage patterns.
Breaking API changes to check
Google lists the same breaking changes for both migrations in the Nov. 2 release notes: function calling and thought_signature validation, thinking_level replacing thinking_budget, and removed sampling parameters and prefilled model turns. The deprecations page points to the Gemini 3.8 Flash migration guide for Pro replacements and the Gemini 3.5 Flash-Lite migration guide for Flash replacements—those guides are where Google documents request-shape changes beyond a one-line model rename.
Who is affected
Access through the API is limited. For gemini-2.5-pro, Google says API access is “limited to existing users who have actively used it in the past.” The Flash deprecations callout adds that access to the 2.5 models is limited to users who have actively used them in the past.
If you never called these endpoints, you may already be unable to select them in Studio and could face API restrictions independent of the Dec. 1 reroute.
gemini-2.5-flash-lite is explicitly not deprecated and remains available through the API until further notice, with no shutdown date announced on the deprecations table.
What developers should do before Dec. 1
Google's own guidance, distilled from the Nov. 2 notes and deprecations callouts:
- Inventory every call site, config file, and agent prompt that still references
gemini-2.5-proorgemini-2.5-flash. - Update Pro workloads to
gemini-3.8-flashand Flash workloads togemini-3.5-flash-litebefore Dec. 1, 2026. - Read the Gemini 3.8 Flash migration guide for Pro and the Gemini 3.5 Flash-Lite migration guide for Flash, paying attention to function calling, thought signatures, thinking levels, sampling parameters, and prefilled turns.
- Re-run integration tests with the new model strings before Dec. 1 instead of relying on automatic routing.
- Review the pricing pages for the destination models.
What it means
Analysis (ours). Pro traffic lands on 3.8 Flash while Flash traffic lands on 3.5 Flash-Lite—different capability and billing profiles. Silent routing protects stragglers but can hide behavioral drift if you skip the migration guides. The “actively used in the past” gate narrows who still touches these endpoints; Dec. 1 mainly matters for legacy strings in scripts and CI. For where Google is steering flagship work, see Gemini 4 Argon Now Available: Price, Access, What Changed.
What to watch
Project-level shutdown notices beyond the Dec. 1 earliest date on the deprecations table, routing behavior for batch jobs that still embed 2.5 IDs, and whether inactive-user limits tighten before Dec. 1. Confirm gemini-2.5-flash-lite stays off the deprecation list if your stack depends on it.