Fast, Efficient AI with Controllable Reasoning
Gemini 2.5 Flash, is now in preview, offering improved reasoning while prioritizing speed and cost efficiency for developers.
Gemini 2.5 Flash delivers major upgrades in reasoning capabilities while prioritizing speed and cost, offering the best price-to-performance ratio among thinking models. Developers can toggle thinking on or off and set thinking budgets to balance quality, cost, and latency for their specific use case.
As a fully hybrid reasoning model, Gemini 2.5 Flash performs a thinking process before responding, allowing it to break down complex tasks and plan responses. It ranks second only to Gemini 2.5 Pro on Hard Prompts in LMArena, making it suitable for reasoning-intensive workloads.
With thinking budget set to 0, Gemini 2.5 Flash maintains the fast speeds of 2.0 Flash while delivering improved performance. This allows developers to get performance gains without latency penalties for use cases that don't require extended reasoning.
Developers can set a thinking budget from 0 to 24,576 tokens to control maximum reasoning tokens. The model automatically decides how much to think based on task complexity, allowing optimization for low-reasoning tasks (simple queries) to high-reasoning tasks (engineering calculations) without overspending.