Google releases Gemini 3 Pro with state-of-the-art reasoning and multimodal capabilities

Google has released Gemini 3 Pro in preview, introducing its most intelligent model to date with significant advancements in reasoning, multimodal understanding, and agentic coding. The release includes the new Gemini 3 Deep Think mode for enhanced performance and is being deployed across Google Search, the Gemini app, AI Studio, Vertex AI, and the Google Antigravity platform.

• Gemini 3 Pro tops the LMArena Leaderboard with a score of 1501 Elo and achieves PhD-level reasoning on Humanity’s Last Exam (37.5%) and GPQA Diamond (91.9%).
• The model sets new state-of-the-art benchmarks in mathematics (23.4% on MathArena Apex), multimodal reasoning (81% on MMMU-Pro), and factual accuracy (72.1% on SimpleQA Verified).
• Gemini 3 Deep Think mode further improves performance, scoring 41.0% on Humanity’s Last Exam, 93.8% on GPQA Diamond, and 45.1% on ARC-AGI-2.
• The model features a 1 million-token context window and is integrated into AI Mode in Search to generate dynamic, interactive visual layouts and simulations.

This release marks the first time Gemini is shipped in Search on day one and aims to provide developers and users with powerful tools for complex problem-solving, learning, and autonomous coding tasks.