Stanford Releases Terminal-Bench-Science: Claude Opus 5 Solves Only ~30%
BenBlaiszik · x · 2026-08-28
Stanford-led community project Terminal-Bench-Science (v0.1) is released to evaluate AI agents on research workflows across life, physical, mathematical, and earth sciences. It features 70 tasks built from real computational workflows with a high-quality bar via expert review. Initial results show Claude Opus 5 solving only about 30% of the tasks.
Related event: Stanford-led Terminal-Bench-Science Debuts; Claude Opus 5 Passes Just 30%(19 posts)→
More from coding & agent
- Dev pushes back on "AI coding means insecure apps": humans always shipped bugs — dreamwieber · 2026-10-15
- Eight non-negotiable principles for building AI that actually learns from experience — tokenbender · 2026-10-15
- Animated teardown of XBOW's Chrome full-chain exploit that won Google's first $250K VRP bounty — moyix · 2026-10-15
- pstack's durable core: 10-20 line skills that distill agent best practices into principles — HankYeomans · 2026-10-15
- 'Children of Elon' forum lets AI agents post and teach each other while humans watch — AiJohnAllen · 2026-10-15
- VoidZero open-sources Void: SDK, CLI and platform to deploy Vite apps on Cloudflare — cnakazawa · 2026-10-15