Skip to content

Gemini 2.5 Pro, Flash and Flash-Lite retire on Vertex AI today

Google shuts down all three Gemini 2.5 models on its Vertex AI platform on 20 October. The Gemini Developer API is unaffected, so the cutoff depends on where you call the model.

Source: Google Cloud

Gemini 2.5 Pro, Gemini 2.5 Flash and Gemini 2.5 Flash-Lite stop working on Vertex AI today, 20 October. Google’s own model lifecycle page lists all three for retirement on this date, and developers calling them by model ID on that platform will start seeing errors rather than responses.

The cutoff is platform-specific, not model-wide. The Gemini Developer API at ai.google.dev has not announced a shutdown date for the same three models, so code calling Gemini 2.5 through the consumer-facing API keeps working. Only Vertex AI, now folded into what Google calls the Gemini Enterprise Agent Platform, is affected. Tuned versions of Gemini 2.5 Flash, including models fine-tuned with supervised tuning or preference tuning, ride on the same retirement date as the base model.

model="gemini-2.5-flash"  # fails on Vertex AI from today
model="gemini-3.5-flash-lite"  # listed as a same-price replacement

Why it matters: if your production code calls Gemini 2.5 through Vertex AI, by model ID, this is the day it breaks, not a warning for later. Check which API surface you’re actually hitting before assuming you’re safe.

The caveat: Google hasn’t published official guidance in this thread on whether tuned Gemini 2.5 Flash models keep functioning after the base model retires, so if you’ve fine-tuned one, don’t assume it survives untested.

Share

Related reading

ML for Web Devs

Tokens, context windows, and why your bill is what it is

The unit of billing for language models is not the request or the word. Once you understand tokens and how a context window fills up, the invoice stops being a surprise.

ML for Web Devs

RAG is just search with extra steps

Retrieval-augmented generation gets written up like an architecture. It is a search query, a string concatenation and one API call, and the search query is the part that decides whether it works.

ML for Web Devs

A prompt is a spec, so write it like one

Prompt engineering has very little to do with magic words. It is the same skill as writing a clear ticket for a capable contractor who will not ask you any questions.

More from The Wire

All briefs

GitLab patches a second CVSS 10.0 RCE in nine days

CVE-2026-110264 is an unauthenticated code execution bug in the Onigmo regex engine, reached through Org-mode file previews. It reaches back to 2014 and catches servers already patched for the 12 October flaw.

via GitLab