Gemini 2.5 Pro and Flash retire on 1 December 2026: what to do before the date
Google shuts down gemini-2.5-pro and gemini-2.5-flash on the Gemini API on 1 December 2026. The recommended replacements are gemini-3.8-flash for 2.5 Pro and gemini-3.5-flash-lite for 2.5 Flash. From that date Google routes requests for the old IDs to these models and bills them at the new models' rates. Your code keeps running on a different model. Re-test your own tasks before the date.
The dates, by platform
| Platform | Model ID | Shutdown | Recommended replacement | Source |
|---|---|---|---|---|
| Gemini APIgemini-2.5-pro | gemini-2.5-pro | 01 Dec 2026 | gemini-3.8-flash | |
| Gemini APIgemini-2.5-flash | gemini-2.5-flash | 01 Dec 2026 | gemini-3.5-flash-lite | |
| Gemini API, preview TTSgemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts | gemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts | 17 Nov 2026 | gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts | |
| Gemini API, Live previewgemini-2.5-flash-native-audio-preview-12-2025 | gemini-2.5-flash-native-audio-preview-12-2025 | 17 Nov 2026 | gemini-3.8-live |
Read from Google's Gemini API deprecations page on 12 November 2026. That page says gemini-2.5-flash-lite is not deprecated. Vertex AI has its own dates: on 12 November 2026 Google's Vertex AI model versions page (Gemini Enterprise Agent Platform) listed a retirement date of 20 October 2026 for gemini-2.5-pro, gemini-2.5-flash and gemini-2.5-flash-lite, while still showing them in its table of available models. If you call Gemini 2.5 through Vertex AI, check that page for your project now. Every other retiring model is in our model retirement calendar.
What happens if you do nothing
- Stable 2.5 IDs keep answering, from a different model. Google's page says that from 1 December 2026 all API requests to gemini-2.5-pro route to gemini-3.8-flash and requests to gemini-2.5-flash route to gemini-3.5-flash-lite. Nothing errors, so nothing alerts you. Answers, tool calls and output format can change silently.
- Your bill changes. Routed requests are billed at the replacement model's rates. Token use per answer can move too. Measure cost per finished task, not price per token.
- Google lists breaking API changes. It links a migration guide for each replacement and asks you to review it before 1 December. Read the one for your target model.
- Pinned IDs hide the switch. Your logs and config will still say 2.5. Teams that pinned an ID to keep behaviour stable lose that stability without a code change.
- Preview models stop. On Google's page, a shut-down model's endpoint is no longer available. The preview TTS and native-audio models above shut down on 17 November 2026; no routing is stated for them.
- Defaults and aliases. If your framework or SDK picks the model for you, check which ID it sends. You may be on 2.5 without knowing it.
Migration checklist
- Find every call site and model ID. Search code, config, environment variables and framework defaults for gemini-2.5. Include batch jobs and evaluation scripts.
- Freeze 20–30 real tasks with expected answers. Take them from production, include the cases that must never go wrong, and write the answers down before any run.
- Run the old and candidate setups on the same tasks. 2.5 today, the named replacement, and one alternative. Same prompts, same tools, same grading.
- Compare cost per finished task and the named tasks that change. Count retries and failures in the cost. List exactly which tasks each setup gets wrong; an average hides them.
- Check the mechanics. Prompts, tool calls, JSON output against your schema, latency and rate limits on the new model.
- Roll out behind a flag. Move a share of traffic, watch the frozen set and production, then switch fully before 1 December.
- Keep the evidence. Store the task set, results and decision. The next retirement needs the same test.
Why averages mislead
In our public test of the Aider coding agent on 27 tasks with hidden tests, Claude Sonnet solved the same 26 tasks as Claude Opus at 49% lower cost per solved task. Claude Haiku was far cheaper but failed 4 more tasks, mostly edge cases. An average hides that; the named failures show whether those 4 tasks matter to you.
That test used Claude models, not Gemini. The method is the same for any switch, including Gemini 2.5 to 3.8 Flash or 3.5 Flash-Lite: same tasks, known answers, cost per finished task, failures by name. Full method and raw data: Aider cost vs quality report.
Test your switch before 1 December
Free pilot: send 20–30 of your tasks with expected answers. We run 3 setups (your Gemini 2.5 setup and two candidates) and send a one-page report: cost per finished task and every task each setup gets wrong.
7-day test, $1,390 fixed: 20 tasks, 3 setups, a signed report in 7 working days with a recommendation.
We work in writing: use the “Write to us” chat on this page or e-mail hello@agenttrustapi.com.
For agents: the calendar is also a JSON file and an Atom feed, each with source and check date.
Questions people ask
When are Gemini 2.5 Pro and Gemini 2.5 Flash retired?
On the Gemini API, gemini-2.5-pro and gemini-2.5-flash shut down on 1 December 2026, per Google's Gemini API deprecations page (checked 12 November 2026). The preview models gemini-2.5-flash-preview-tts, gemini-2.5-pro-preview-tts and gemini-2.5-flash-native-audio-preview-12-2025 shut down on 17 November 2026.
What replaces Gemini 2.5 Pro and Gemini 2.5 Flash?
Google recommends gemini-3.8-flash for gemini-2.5-pro and gemini-3.5-flash-lite for gemini-2.5-flash. Test at least one alternative on your own tasks too; the named replacement is not always the best fit for cost or quality.
Will my Gemini 2.5 requests fail after 1 December?
Not for the stable IDs on the Gemini API. Google says requests to gemini-2.5-pro and gemini-2.5-flash route automatically to the replacement models and are billed at their rates. The call succeeds, but a different model answers, so quality and cost can change without any error.
Is gemini-2.5-flash-lite retired too?
Not on the Gemini API: Google's deprecations page says it is not deprecated and remains available until further notice. Vertex AI dates differ; see the next answer.
Are the Vertex AI dates the same?
No. Vertex AI (Gemini Enterprise Agent Platform) keeps its own schedule. On 12 November 2026 the Vertex AI model versions page listed 20 October 2026 for the three Gemini 2.5 text models, with gemini-3.8-flash, gemini-3.5-flash or gemini-3.5-flash-lite as replacements. Check that page for your project.
How do I test the switch before the deadline?
Freeze 20–30 real tasks with expected answers, run your Gemini 2.5 setup and the candidates on the same set, compare cost per finished task and the named tasks that change, check JSON output, tool calls and limits, and roll out behind a flag. We run this as a free pilot, or as a fixed 7-day test for $1,390.
Questions or corrections: hello@agenttrustapi.com.