Google deprecated Gemini 2.5 Pro and Gemini 2.5 Flash on November 2, 2026. On December 1, 2026, API requests to gemini-2.5-pro route automatically to gemini-3.8-flash, and gemini-2.5-flash routes to gemini-3.5-flash-lite, each billed at the new model’s rates. Update model strings and fix breaking parameters before then.
Last updated: November 8, 2026
Table of contents
- What is changing
- What it will cost
- Breaking API changes
- How to migrate step by step
- Choosing a target model
- Frequently asked questions
- Sources
What is changing
The Gemini API changelog lists two deprecations dated November 2, 2026. Both models were removed from the model picker in Google AI Studio that day.
The deprecations page adds that API access is limited to users who have actively used the models in the past. It also says gemini-2.5-flash-lite is not deprecated and remains available until further notice. Older preview versions of both models are also listed as deprecated.
| Model | Shutdown date listed | Automatic routing from December 1, 2026 |
|---|---|---|
| gemini-2.5-pro | December 1, 2026 | gemini-3.8-flash, billed at Gemini 3.8 Flash rates |
| gemini-2.5-flash | December 1, 2026 | gemini-3.5-flash-lite, billed at Gemini 3.5 Flash-Lite rates |
| gemini-2.5-flash-lite | Not deprecated | No change announced |
Automatic routing means your application keeps working, but the model answering is different. Outputs, latency and cost can change without any code edit, so treat December 1 as a forced release date for your prompts.
What it will cost
The pricing page lists paid-tier prices per million tokens. The figures below are for standard text input and output.
| Model | Input | Output (including thinking) |
|---|---|---|
| Gemini 2.5 Pro (prompts up to 200k tokens) | $1.25 | $10.00 |
| Gemini 3.8 Flash, through Dec 31, 2026 | $0.75 | $3.75 |
| Gemini 3.8 Flash, from Jan 1, 2027 | $1.50 | $7.50 |
| Gemini 2.5 Flash (text, image, video input) | $0.30 | $2.50 |
| Gemini 3.5 Flash-Lite | $0.30 | $2.50 |
For 2.5 Pro users, 3.8 Flash is cheaper until the end of the year, then costs more on input than the old Pro price. The 3.8 Flash migration guide describes the current prices as introductory. Plan the January 1, 2027 step before it arrives. For other Gemini price questions, see our Gemini API pricing guide.
Breaking API changes
The changelog warns that moving to the newer models brings changes beyond the model name. The Gemini 3.8 Flash guide repeats them in its migration section:
- Thinking control: thinking_level replaces thinking_budget. On 3.8 Flash, choose low, medium (default) or high. The value minimal is not supported there.
- Sampling parameters: temperature, top_p and top_k are deprecated and ignored, and future model generations return an error. candidate_count is unsupported in Gemini 3 and later.
- Prefilled turns: If the last non-empty turn is a model turn, the API returns a 400 error. Use previous_interaction_id for server-side history.
- Function calling: Place multimodal assets inside the response payload, and with generateContent include call_id and name in every FunctionResponse. Thought signatures must be preserved.
The Gemini 3.6 Flash and 3.5 Flash-Lite guide describes the same parameter rules and recommends leaving thinking_level at minimal for high-volume extraction and routing, and medium or high for tool-calling subagents.
How to migrate step by step
- Search your code, configuration and notebooks for gemini-2.5-pro and gemini-2.5-flash, including preview model names.
- Replace gemini-2.5-pro with gemini-3.8-flash and gemini-2.5-flash with gemini-3.5-flash-lite, or pick another model from the guides below.
- Replace thinking_budget with thinking_level and remove temperature, top_p, top_k and candidate_count.
- Remove any request that ends with a prefilled model turn, and move conversation state to previous_interaction_id if you can.
- Update your SDK, check that function responses carry call_id and name, and keep thought signatures intact.
- Run your evaluation set on the new model, compare quality and cost, and adjust prompts before December 1.
Choosing a target model
The automatic targets are defaults, not requirements. The deprecations page names gemini-3.8-flash and gemini-3.5-flash-lite as the replacements, and the pricing page describes 3.8 Flash as built for long-horizon software engineering, autonomous agents and enterprise workflows. Gemini 3.5 Flash-Lite is described as cost-efficient and tuned for high-volume agentic tasks, translation and simple data processing.
If you already build on newer Google tooling, our guides to the Gemini CLI 0.67 release, the Gemini API CLI and the Computer Use tool show current options, and our Gemini Omni 1.2 Flash guide and voice remixing guide cover the other November API changes.
Frequently asked questions
When is Gemini 2.5 Pro being deprecated?
Google deprecated gemini-2.5-pro on November 2, 2026, and it is no longer selectable in Google AI Studio. API access is limited to existing users. On December 1, 2026, API requests route automatically to gemini-3.8-flash.
What replaces Gemini 2.5 Flash?
Effective December 1, 2026, requests to gemini-2.5-flash route automatically to gemini-3.5-flash-lite and are billed at Gemini 3.5 Flash-Lite rates. The deprecations page also says gemini-2.5-flash-lite is not deprecated and stays available until further notice.
Will my bill change after December 1?
Yes. Requests are billed at the replacement model’s rates. Gemini 3.8 Flash costs $0.75 input and $3.75 output per million tokens through December 31, 2026, then $1.50 and $7.50 from January 1, 2027, according to the pricing page.
Do I have to change my code?
Routing keeps requests working, but Google lists breaking changes. thinking_level replaces thinking_budget, the temperature, top_p and top_k parameters are removed, and prefilled model turns are no longer accepted. Testing before December 1 avoids surprises.
Why does 3.8 Flash not accept minimal thinking?
The Gemini 3.8 Flash guide states that minimal is not supported. Choose low, medium (the default) or high. Gemini 3.5 Flash-Lite keeps minimal as its default level.
Sources
- Gemini API changelog (entry of November 2, 2026)
- Gemini API deprecations
- Gemini 3.8 Flash guide and migration steps
- Gemini API pricing
Image: Google Gemini logo via Wikimedia Commons, public domain. Gemini is a trademark of Google. FreeAnswers is not affiliated with or endorsed by Google.
