Gemini 3 sets new reasoning records

Google's Gemini 3 Pro tops LMArena and posts ~37.5% on Humanity's Last Exam without tools and 31% on ARC-AGI-2.

What happened and why it matters

Gemini 3 showed a large step on the hardest reasoning benchmarks of 2025, and its Deep Think mode pushed ARC-AGI-2 higher still — several-fold improvements in under a year on benchmarks designed to resist progress.

Sources

• A new era of intelligence with Gemini 3 ↗ Google Blog · 18 Nov 2025