Google has deprecated Gemini 2.5 Pro and Flash. Existing API calls are scheduled to route to different replacements on 1 December, with replacement-model billing. A successful request is not proof that the application still behaves the same.
What changes, and when
Google’s 2 November Gemini API release notes announce deprecation of gemini-2.5-pro and gemini-2.5-flash and their removal from the Google AI Studio model selector. From 1 December 2026, requests to the Pro ID are scheduled to route to gemini-3.8-flash; requests to the Flash ID to gemini-3.5-flash-lite. Google says billing will use the replacement models’ rates. This is an API lifecycle change, not a claim that every Gemini app user must change plans.
Do not confuse the three 2.5 IDs
The deprecation schedule distinguishes Pro and Flash from gemini-2.5-flash-lite, for which it lists no announced shutdown date. It also says access to the 2.5 models is limited to previous active users. Confirm the literal model string in each application rather than treating every service labelled “Gemini” as the same route. A third-party integration needs its own supplier confirmation; this report does not verify access for every account or region.
A model-string edit is not the whole migration
The 3.8 Flash migration guide replaces thinking_budget with thinking_level and lists changes to sampling parameters, prefilled turns and function-call handling. Review the guide for the API and SDK you actually use. Our suggested acceptance check is a permitted, known-answer document task followed by a tool-error case. Inspect the returned fields and the downstream decision, not just whether the server returns success. HubAI has not benchmarked these replacement routes for this report.
Budget December and January separately
For Standard paid Gemini 3.8 Flash usage, Google lists US$0.75 per million input tokens and US$3.75 per million output tokens, including thinking, through 31 December 2026. The listed rates from 1 January 2027 are US$1.50 and US$7.50 respectively. These are usage prices, not monthly subscriptions or a complete application quote. Other tiers, caching and tools have separate terms. Google’s pricing table distinguishes product-improvement use on Free from the paid Standard route. Check the applicable terms and logging settings before supplying confidential material; paid access alone is not an organisation’s approval.
Leave an auditable decision
Name the migration owner, record the old and replacement IDs, save the request configuration and agree a stop condition. Include retries, reviewer time and rejected outputs in the cost comparison. If a deadline forces the old route to change, a safer fallback may be a manual queue rather than an untested automatic substitute. Use the Model Pilot guide below to prepare the acceptance record; do not describe a small trial as proof of general reliability.
HUBAI VIEWInventory the actual model IDs, test request compatibility and approve the complete workload cost before the scheduled change.
Buyer decision signal: Google release notes · 2 November announcement, 1 December routing change
What to verify next
1Find every literal model ID and integration owner.
2Check the correct replacement and API compatibility.
3Compare December and January workload costs.
4Approve data handling, exception review and a safe fallback.
Read the evidence
Capabilities, availability and prices can change. HubAI keeps analysis separate from the underlying official material.
01Google · Gemini API release notes, 2 November 2026Open source ↗02Google · Gemini deprecation scheduleOpen source ↗03Google · Gemini 3.8 Flash migration checklistOpen source ↗04Google · Gemini API pricing by usage tierOpen source ↗
Models
Models