gemini-2.5-flash on Gemini API is routed to gemini-3.5-flash-lite on December 1, 2026

AI model

Date
(UTC)
Replacement
gemini-3.5-flash-lite
Platform
Gemini API
Provider
Google
Last checked

gemini-2.5-flash

Routes to gemini-3.5-flash-lite on Gemini API

In its place: gemini-3.5-flash-lite

since.dev/calendar/google/gemini-2.5-flash/gemini

What changes

Requests keep working, answered by gemini-3.5-flash-lite.

What works differently on gemini-3.5-flash-lite

From its documentation. Moving to it is more than a new model ID.

  • The default thinking level is minimal, so a request that does not set thinking_level runs at minimal. To keep the old behavior: for autonomous subagents with tool calls, code execution, or multi-step reasoning, set thinking_level to "medium" or "high". Source
  • temperature, top_p, and top_k are deprecated; the API ignores these parameters and returns an error in future model generations. To keep the old behavior: remove these parameters from all requests. Source
  • Prefilled model turns are no longer supported; if the last non-empty turn in the request is a model turn, the API returns a 400 error. To keep the old behavior: instead of prefilling a model turn, use system_instruction or Structured outputs. Source
  • candidate_count is unsupported in Gemini 3 and later. To keep the old behavior: remove candidate_count from requests. Source

Source

aistudio.google.com

gemini-2.5-flash | June 17, 2025 | December 1, 2026 | gemini-3.5-flash-lite (gemini-2.5-flash is no longer available in Google AI Studio; effective December 1, 2026, API requests to gemini-2.5-flash will automatically route to gemini-3.5-flash-lite)

Data license: CC BY 4.0. From Google's documentation. Quotes are shortened and dates restated.

Does your code use it?

Paste a public repository. Since.dev analyzes it and shows the files and lines these dates hit.

Public repository. No sign in.

A warm wave of orange and red light over deep blue

Put your code on continuous maintenance.

Scan a public repository now, or give your agent one line.

See how out of date your code is.

Any public GitHub repository. No sign-in.

Tell your agent what changed since it was trained.

claude mcp add --transport http since https://api.since.dev/mcp

Codex, Cursor and any MCP client use https://api.since.dev/mcp.