Google Extends Gemini 2.5 Retirement on Vertex AI to October 2026
Google has pushed the Vertex AI retirement date for Gemini 2.5 Pro, Flash, and Flash-Lite to October 16, 2026, giving developers more time to migrate to Gemini 3.x.
Google has updated the retirement dates for Gemini 2.5 Pro, Gemini 2.5 Flash, and Gemini 2.5 Flash-Lite on Vertex AI, pushing the cutoff from June 2026 to October 16, 2026. If you have production workloads running on any of these models, you now have a longer runway to migrate, but the clock is still running.
What exactly is changing
Three Gemini 2.5 models on Vertex AI are affected:
- Gemini 2.5 Pro
- Gemini 2.5 Flash
- Gemini 2.5 Flash-Lite
All three now carry an October 16, 2026 retirement date on Vertex AI specifically. One month before that date, Google will block new access to these models for online inference, batch inference, and tuning. After October 16, any API call referencing a retired model ID will return a 404 error, and the model is gone.
It is worth noting that this extension applies to Vertex AI only. On AI Studio and the Gemini API, these models are still scheduled for deprecation on June 17, 2026. If you use both surfaces, you need to track both timelines separately.
Also worth flagging: Google’s policy describes listed shutdown dates as “the earliest possible dates,” so October 16 could shift again. That said, planning to that date is the only sensible approach.
The path forward: migrating to Gemini 3.x
The Gemini 3 family has been rolling out since late 2025, and the current generation looks like this:
- Gemini 3.1 Pro launched in February 2026. It is Google’s most capable reasoning model in the series, supports a 1 million token context window, and handles text, audio, images, video, PDFs, and code repositories. It is available in preview in Model Garden and is the most direct replacement for Gemini 2.5 Pro.
- Gemini 3.1 Flash is built for speed with strong reasoning. If you are using Gemini 2.5 Flash today, this is the natural migration target.
- Gemini 3.1 Flash-Lite is in public preview and is Google’s most cost-efficient option, optimised for high-volume, latency-sensitive workloads.
The API-level change is relatively contained. The model ID is the main thing that changes in your requests. However, models do behave differently, and prompts tuned carefully to 2.5 outputs can drift when you move to 3.x. Factor in time to re-run your evaluation set before cutting over in production.
What this means for your planning
A few things worth keeping on your radar as you plan the migration.
Provisioned Throughput users need extra lead time. If you have purchased Provisioned Throughput on Vertex AI, you will need to update your PT assignment to point at the new model endpoints before the retirement date. Google recommends submitting those requests about a month in advance to allow for approval processing.
The SDK situation is worth addressing now. Google has noted that Vertex AI SDK releases after June 2026 will not support Gemini models. New Gemini features are only available through the Gen AI SDK. If your codebase still uses the older Vertex AI SDK, that is a migration worth doing sooner rather than treating it as an afterthought on the way to Gemini 3.
Fine-tuned endpoints have their own shutdown timing. If you have supervised fine-tuned endpoints built on Gemini 2.5 Pro, those will be shut down the day after the October 16 retirement date. Plan accordingly if fine-tuning is part of your setup.
This is not a one-and-done migration. Gemini 2.5 Flash already carries an October 16, 2026 shutdown date, which means teams migrating from Gemini 2.0 Flash to 2.5 Flash as an interim step will need to move again within a short window. If you are at that stage, it is worth considering jumping straight to Gemini 3.1 Flash and skipping the intermediate hop.
The broader context
Google’s model lifecycle policy gives you some predictability here. A retirement date is announced once the next stable version of a model is publicly released. One month before retirement, new access is blocked. After the retirement date, the model is no longer accessible.
The October 16 date also has a conditional element. Google has said that a confirmed discontinuation date will be set once Gemini 3 reaches General Availability, and that they will provide at least six months of notice from that point. So if Gemini 3 GA is announced, the clock on a firmer deadline starts there. Keep an eye on the Vertex AI release notes for any updates.
The short version
If you are running Gemini 2.5 Pro, Flash, or Flash-Lite on Vertex AI, you have until October 16, 2026 to migrate to a Gemini 3.x equivalent. The extension from June is welcome breathing room, but it is not an indefinite delay. Start your evaluation work now, sort out your SDK version if needed, and if you are on Provisioned Throughput, get that request in well ahead of the deadline.