Google is rolling out an early preview of Gemini 2.5 Flash, building on the foundation of 2.0 Flash. This new version offers significant enhancements in reasoning abilities while maintaining speed and cost-effectiveness.
Key Highlights:
- It’s the first fully hybrid reasoning model, allowing developers to enable or disable “thinking” and set budgets.
- Setting “thinking off” maintains the speed of 2.0 Flash while improving performance.
- The model can perform a “thinking” process to better understand prompts and plan responses.
- Gemini 2.5 Flash performs strongly on Hard Prompts in LMArena.
- It is cost-efficient.
- Developers can control the “thinking budget” to manage quality, cost, and latency.
- It’s available in preview through the Gemini API via Google AI Studio and Vertex AI, and in the Gemini app.
https://deepmind.google/discover/blog/introducing-gemini-2-5-flash/