Google Launches Gemini 3 AI Model with Record-Breaking 1501 Elo Score and Deep Think Reasoning Mode
Summary
Google launches Gemini 3 AI model with record-breaking 1501 Elo score, featuring new Deep Think reasoning mode that achieves 45.1% on ARC-AGI-2 benchmarks and rolls out today across Google Search, Gemini app, and new Antigravity development platform.
Key Points
- Google launches Gemini 3, its most intelligent AI model to date, featuring breakthrough performance across reasoning, multimodal understanding, and coding benchmarks with a top LMArena score of 1501 Elo
- Gemini 3 Deep Think mode delivers enhanced reasoning capabilities for complex problems, achieving 45.1% on ARC-AGI-2 and 93.8% on GPQA Diamond benchmarks
- Gemini 3 becomes available today across Google products including Search AI Mode, Gemini app, AI Studio, and the new Google Antigravity agentic development platform