A new era of intelligence with Gemini 3

Google is releasing Gemini 3, its most intelligent model yet, which combines advanced reasoning with stronger multimodal understanding. The model is rolling out across AI Mode in Search, the Gemini app, AI Studio, Vertex AI and other developer tools, while Gemini 3 Deep Think is entering safety testing before availability to Google AI Ultra subscribers.

Gemini 3 Pro outperforms previous generations on major benchmarks, topping the LMArena leaderboard with a 1501 Elo score and advancing performance in mathematics, multimodal tasks and factual accuracy. The team says the model delivers concise, nuanced responses and better grasps context and intent, reducing the need for extensive prompting.

The model is positioned to help people learn, build and plan: it can translate handwritten recipes, synthesize long lectures into interactive study tools, analyze videos to suggest training plans, and generate code for high‑fidelity visualizations.

gemini 3, google, ai mode, gemini app, ai studio, vertex ai, deep think, elo score, multimodal, code generation