ConjectureBench Released as Frontier AI Models Crack Decades-Old Math Problems

Bespoke Labs introduced ConjectureBench, featuring 15,000 open mathematical problems for AI agents. The benchmark demonstrates that frontier models can now solve decades-old problems, such as overturning the 80-year-old Jacobian conjecture using Claude 3.5.

2026-08-01 ~ 2026-08-01 · 2 related posts

• Claude 3.5 Helps Disprove 80-Year-Old Jacobian Conjecture — gregd_nlp · 2026-08-01
• ConjectureBench: 15k open math problems for AI agents, frontier models crack decades-old conjectures — RexDouglass · 2026-08-01