Science's Last Exam: 91-task autonomous research benchmark, best model scores 67, all fail end-to-end
AccBalanced · x · 2026-11-06
The AItonomy Foundation launches Science's Last Exam (SLE), a benchmark for autonomous end-to-end scientific research: 91 tasks (113 per the official post) across 9 disciplines, crafted with experts from 34 institutions including Harvard, MIT and Stanford. Agents get real research goals and real tools (lab data, software) and are graded against hidden expert checks covering protocol planning, experiment observation, scientific computing and software operation. Reported scores: Claude Opus 5 at 67.0, GPT-6.1 Sol at 64.25, Kimi K3.1 Pro at 58.25; the official post says the best of 8 frontier models hit 37.5/100 and all scored 0 on true end-to-end research. Everything opens Nov 6; Claude Opus 5.5 and GPT-6 Astra are pending bio safeguards.
Related event: New Science Benchmark Exposes Weakness of Frontier AI Models in Research(3 posts)→
More from Models
- ValsAI claims critical safety flaw in frontier lab models enabling high-stakes misuse — rickasaurus · 2026-11-06
- Follow-up: gradient flow lets residual states encode information for far-future predictions — mlpowered · 2026-11-06
- Benchmarking which AI lies best, using a roleplaying game — ownworldman · 2026-11-06
- Kimi K3.1 Pro gets Day-0 TPU support: serving a 2.8T-param MoE with llm-d on TPUs and GPUs — vllm_project · 2026-11-06
- Google AI Studio spotted offering $60 in API credits on some accounts — ygsbhardwaj · 2026-11-06
- Zhipu chief scientist reveals GLM-5.5 training pipeline: 3 sizes, 6 internal data engines — AccBalanced · 2026-11-06