Gemini 2.5 Pro Retirement: How to Migrate Before Oct 20
The short answer
If you call Gemini 2.5 through Vertex AI, you have until October 20, 2026 to switch model IDs; Google’s recommended replacement for Gemini 2.5 Pro and 2.5 Flash is Gemini 3.8 Flash. On the Gemini Developer API the stable 2.5 models have no shutdown date yet, but access is restricted to existing users, so plan the move anyway. Replacements from Google’s own lifecycle tables, with Gemini API paid-tier prices per million tokens (input / output) as of October 14, 2026:
| Retiring model | Vertex AI retirement | Google’s recommended replacements | Old price | Replacement price |
|---|---|---|---|---|
gemini-2.5-pro | Oct 20, 2026 | gemini-3.8-flash or gemini-3.5-flash | $1.25 / $10 (≤200K) · $2.50 / $15 (>200K) | 3.8 Flash $0.75 / $3.75 to Dec 31, then $1.50 / $7.50 |
gemini-2.5-flash | Oct 20, 2026 | gemini-3.8-flash, gemini-3.5-flash-lite or gemini-3.1-flash-lite | $0.30 / $2.50 | 3.5 Flash-Lite $0.30 / $2.50 · 3.1 Flash-Lite $0.25 / $1.50 |
gemini-2.5-flash-lite | Oct 20, 2026 | gemini-3.8-flash, gemini-3.1-flash-lite or Gemma 4 | — | 3.1 Flash-Lite $0.25 / $1.50 |
Two more dates on Vertex AI: gemini-3.6-flash retires November 19, 2026 and gemini-3.7-flash January 28, 2027, both replaced by 3.8 Flash. On the Gemini API, gemini-3.7-flash was deprecated on October 8, 2026 and its requests are routed to 3.8 Flash automatically.
Which replacement to pick
Gemini 2.5 Pro users: start with Gemini 3.8 Flash. Google positions it as its most capable Flash model for long-horizon coding and agents; it costs less than 2.5 Pro until the end of 2026 and about the same input price from January 2027. Budget at the January rate ($1.50/$7.50). If 3.8 Flash fails your hardest evals, test Gemini 4 Argon (GA since October 13, 2026; $2/$10 until November 30, then $4/$20) — see Gemini 4 Argon vs Gemini 3.8 Flash.
High-volume Gemini 2.5 Flash users: Flash-Lite first. Gemini 3.1 Flash-Lite is cheaper than 2.5 Flash on both input and output; Gemini 3.5 Flash-Lite matches its price exactly. Move to 3.8 Flash only for the share of traffic that needs more reasoning.
Gemini 2.5 Flash-Lite users: 3.1 Flash-Lite, or self-host Gemma 4 if you want to stop paying per token. Google lists Gemma 4 as a replacement for 2.5 Flash-Lite.
Migration checklist
- Find every call site. Search code and config for
gemini-2.5-; include batch jobs, Cloud Functions, and model IDs stored in databases or prompt-management tools. - Pin the new model ID explicitly. Avoid aliases that silently move.
- Re-run your eval set on the replacement at the same temperature and reasoning settings. Newer Gemini models think by default; check output-token counts, because thinking tokens are billed as output.
- Re-tune prompts that relied on 2.5 quirks (verbosity, JSON formatting). Use structured output instead of format instructions where possible.
- Recalculate cost with real token counts, including the January 1, 2027 Flash price change. See current API prices.
- Ship before October 20 with a fallback model configured, then remove the 2.5 IDs.
For a broader look at Google’s lineup, see Gemini 3.8 Flash vs DeepSeek V4.1 Flash vs V4 Pro and our cheap LLMs for high-volume agents ranking.
Last verified: October 14, 2026.