Gemini 2.5 Flash
Gemini 2.5 Flash preview is the first fully hybrid reasoning Gemini, with thinking switchable on or off and a developer-set thinking budget of 0 to 24,576 tokens.
Brings the 2.5 reasoning recipe to the cheap Flash tier and exposes cost control through a thinking budget. Reached GA 2025-06-17; an adaptive-thinking update followed 2025-05-20.
- Date
- Thursday, 17 April 2025
- Lab
- Google DeepMind
- Kind
- model
- Access
- closed API
Figures
| Measure | Value | Measured by |
|---|---|---|
| Thinking budget range | 0 to 24,576 tokens developer-controlled | company |
Sources
- developers.googleblog.com/en/start-building-with-gemini-25-flash/
- ai.google.dev/gemini-api/docs/changelog
This record was checked against its sources on 6 October 2026. How we check