Google DeepMind has released an AI system called "AI Co-Mathematician" designed to collaborate with human mathematicians on complex problems. This system, built on Gemini 3.1 Pro, achieved a new state-of-the-art score of 48% on the challenging FrontierMath Tier 4 benchmark, significantly outperforming existing models like GPT-5.5 Pro. The AI functions as an asynchronous workspace with a coordinator agent that breaks down tasks, manages parallel research streams, and persistently stores failed hypotheses, mirroring workflows seen in software development. AI
IMPACT This system demonstrates a new paradigm for AI collaboration in research, potentially accelerating discoveries in complex fields like mathematics.
RANK_REASON The cluster describes a new AI system for mathematical research and its performance on a specialized benchmark, including its use in solving a previously unsolved problem.
- AI Co-Mathematician
- Alex Davies
- Claude Opus 4.6
- Claude Opus 4.7
- Daniel M. Roy
- Daniel Zheng
- Epoch AI
- FrontierMath Tier 4
- Gemini 3.1 Pro
- Google DeepMind
- GPT-5.4 Pro
- GPT-5.5 Pro
- Kourovka Notebook
- Marc Lackenby
- Pushmeet Kohli
- Claude Code
- Oxford
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
