New benchmark measures AI teammates after 10,000 OpenAI agents cracked a $1M problem
james_y_zou · x · 2026-10-20
Sean Wu's team introduced the Collaborative Intelligence Benchmark, measuring how well AI models work as teammates. Context: last month 10,000 OpenAI agents jointly cracked a $1M Millennium Prize problem, and OpenAI models were also found to be the best overall teammates.
Key observation: agents are getting smart, but not cooperative—this benchmark targets exactly that gap.
More from Models
- reuF2F benchmark: 244 REU math problems formalized in Lean, best model solves only 109 — wellecks · 2026-10-20
- Andrew Ng: AI math breakthroughs are largely not from mathematicians uploading ideas — andrewgwils · 2026-10-20
- Pangram 4.1 fixes AI-detection false positives after community pushback — GolinoHudson · 2026-10-20
- New models rumored: Gemini Carbon 400% cheaper, Fable 5.5, Terra at $5/1M — bindureddy · 2026-10-20
- Rumor: OpenAI Shelved Unreleased Model Over Misalignment Concerns — haider1 · 2026-10-20
- Multiple leaks point to a new open model tomorrow that fits on a single DGX Spark — ChrisUniverse · 2026-10-20