We’re introducing an upgraded preview of Gemini 2.5 Pro, our most intelligent model yet. Building on the version we released in May and showed at I/O, this model will be the generally available, stable version starting in a couple of weeks, ready for enterprise-scale applications.
The latest 2.5 Pro reflects a 24-point Elo score jump on LMArena, maintaining its lead on the leaderboard at 1470, and a 35-point Elo jump to lead on WebDevArena at 1443. It continues to excel at coding, leading on difficult coding benchmarks like Aider Polyglot. It also shows top-tier performance on GPQA and Humanity’s Last Exam (HLE), highly challenging benchmarks that evaluate a model’s math, science, knowledge and reasoning capabilities.
We also addressed feedback from our previous 2.5 Pro release, improving its style and structure — it can be more creative with better-formatted responses.
Developers can start building with the upgraded preview of 2.5 Pro in the Gemini API via Google AI Studio and Vertex AI — where we’ve also added thinking budgets to give developers more control over cost and latency. It’s also rolling out today in the Gemini app.