Complete Guide to Google Gemini Model Version Deprecation and Update Schedules
A comprehensive overview of Google's latest AI Gemini release status and deprecation schedules from 3.1 to the legacy 1.0-2.5 series. A must-read model migration guide for API developers and practitioners.
Google's core AI models are currently the Gemini series, with continuous version upgrades (1.0 -> 1.5 -> 2.0 -> 2.5 -> 3.x) and a rapid-cycle policy of releasing new versions and deprecating old ones.
Here is a summary of the major release and deprecation/shutdown schedules for Google models as confirmed in March 2026.
- Gemini 3.x Series (Latest Generation)
This is Google's current latest model lineup.
-
Gemini 3.1 Pro Preview (the current model): The newest preview model released in early 2026, currently actively in service.
-
Gemini 3 Pro Preview: Was active from late 2025 to early 2026, but was fully shut down as of March 9, 2026.
- Gemini 2.5 and 2.0 Series (Mainstream and Transitional)
Currently the most widely used in production environments, but deprecation schedules are being set as the transition to the 3.x generation proceeds.
-
Gemini 2.5 Flash / Pro (Stable): Scheduled for deprecation on June 17, 2026.
-
Gemini 2.5 Flash-Lite Preview (09-2025): Scheduled for shutdown on March 31, 2026.
-
Gemini 2.0 Flash / Flash-Lite: Officially retiring on June 1, 2026.
- Gemini 1.0 and 1.5 Series (Legacy Generation)
-
Gemini 1.0 (Pro/Ultra): Announced in late 2023 and served as the flagship in early 2024, but support has now ended for most services and APIs, or it has been fully classified as legacy.
-
Gemini 1.5 (Pro/Flash): The flagship models of mid-2024, now being phased out of API services with the release of 2.x and 3.x.
- PaLM 2 and Other Models (Legacy)
- PaLM 2 (Bison, etc.): Google's flagship model before Gemini, mainly used in 2023. After the Gemini launch, support was mostly ended on Vertex AI and other platforms, with forced migration to Gemini.
Google Model Support Policy Summary
Google shuts down Preview models very quickly — within months of a new version's release — while Stable models are given a grace period of approximately one year before deprecation. API users must continuously migrate to the latest version.
Get new posts by email ✉️
We'll notify you when new posts are published