Gemini 2.5 API Ends Dec 1 — Free Tier Switch
If your code still calls gemini-2.5-flash or gemini-2.5-pro, it has three weeks left. Google deprecated both models on 2 November 2026, and from 1 December your requests will be silently rerouted to newer models — and billed at those models' rates. The rerouting sounds harmless. It isn't: Google lists breaking API changes alongside it, so code that worked on 30 November can fail on 1 December. The fix is a one-line model-string change, but only if you test it first.
What exactly is shutting down
Google's AI Studio changelog entry for 2 November says gemini-2.5-flash "is deprecated and is no longer available to select in Google AI Studio." The same applies to gemini-2.5-pro. The 2.5 models were already restricted before this: since 18 September 2026, Google has served them only to users who had actively used them in the past, and new projects were told to use the latest models instead. Deprecation is the next step — the models stop being served, with 1 December as the date Google gives.
One caution on that date: Google's deprecations page notes that shutdown dates are "the earliest possible dates on which a model might be retired." Treat 1 December as the deadline, not as a date that might slip in your favour.
What happens to your code on December 1
From 1 December 2026, all API requests to gemini-2.5-flash will automatically route to gemini-3.5-flash-lite and be billed at Gemini 3.5 Flash-Lite rates. Requests to gemini-2.5-pro route to gemini-3.8-flash and bill at Gemini 3.8 Flash rates. Your API key keeps working and your requests keep getting answers — from a different model, at a different price.
That is the trap for free-tier users. The reroute keeps your app alive while quietly changing what it costs and how it behaves. If your code was tuned around 2.5 Flash's responses, the replacement model will answer differently, and if you were relying on free-tier pricing, the new model's rates apply. Google's own instruction is short: update your model string before 1 December.
The breaking changes to test now
Do not read "automatically route" as "nothing to do." Google warns of breaking API changes in the new models, including function calling and thought_signature validation, thinking_level replacing thinking_budget, and removed sampling parameters and prefilled model turns. If your 2.5 code uses function calling, thinking_budget, or sampling parameters, test it against the new model now rather than on 1 December.
The practical checklist is short: swap the model string to gemini-3.8-flash (or gemini-3.5-flash-lite if you were on 2.5 Flash and want the cheapest path), run your function-calling flows, and check any code that sets thinking budgets or sampling parameters. Gemini 3.8 Flash went generally available on 2 September 2026, and Google describes it as its most intelligent Flash model — it is the replacement Google itself points new projects to, so it is the one to test against.
What to switch to on the free tier
The free tier survives the shutdown. Gemini 3.8 Flash is on Google's free tier, with input listed as "Free of charge." The catch is that Google no longer publishes free-tier rate limits in a per-model table — your limits now depend on your usage tier and are shown only in your own AI Studio console. The old "15 RPM, 1M tokens a day" figure still quoted around the web is no longer confirmable from a primary source, so check your own console instead of trusting a roundup.
For most free-tier users the switch is: 2.5 Flash users move to 3.5 Flash-Lite, 2.5 Pro users move to 3.8 Flash. Both stay free within the limits your project shows in AI Studio. If you would rather skip API keys entirely, our free coding agents that need no API key roundup covers tools that work without one — and the full free API directory lists the other providers whose free tiers are still live.
What it costs if you outgrow the free tier
If your usage exceeds the free tier, Gemini 3.8 Flash on the paid tier costs $0.75 per million input tokens and $3.75 per million output tokens through 31 December 2026, then rises to $1.50 and $7.50 from 1 January 2027. Those are Google's published standard rates, re-checked against the pricing page on 7 November 2026. Budget for the January jump now if you plan to stay on 3.8 Flash past the new year — the same price rise applies to 3.6 Flash.
FAQ
Do I have to do anything if my app still works?
Yes. Working today is not the test — 1 December is. Change the model string to gemini-3.8-flash or gemini-3.5-flash-lite and run your full flow once before the deadline, especially anything using function calling.
Will my API key stop working on December 1?
No. The key keeps working; the model behind the old string changes. That is exactly why the deadline is easy to miss — nothing visibly breaks on the day, your calls just get answered by a different model at a different price.
I never used Gemini 2.5. Does this affect me?
No. If your code already uses 3.5 Flash-Lite, 3.8 Flash, or another current model, nothing changes. The shutdown only affects the gemini-2.5-flash and gemini-2.5-pro model strings.
Is the Gemini free tier still free after the switch?
Yes — 3.8 Flash input is listed as "Free of charge" on the free tier. The limits are per-project and shown only in your AI Studio console, so check there rather than relying on a published table.
Sources: Google AI Studio changelog and deprecations page, via the 7 November 2026 verification in StackWrite's free AI API guide. Model strings, dates, and prices above are as stated there; free-tier request limits are console-only and vary by project.