Gemini 2.5 Pro leads major leaderboards

Google's first "thinking" Gemini tops LMArena and posts state-of-the-art maths and science scores.

What happened and why it matters

Gemini 2.5 Pro combined a 1M-token context window with built-in reasoning, scoring ~18.8% on Humanity's Last Exam without tools.

Sources

• Gemini 2.5: Our most intelligent AI model ↗ Google Blog · 25 Mar 2025