Agents’ Last Exam (ALE) is a UC Berkeley RDI benchmark (paper, arXiv:2606.05405; leaderboard) of 1,000-plus long-horizon, economically valuable professional tasks with verifiable outcomes, organized into 55 sub-fields across 13 industry clusters and built with 250-plus industry experts; on its hardest tier the average full pass rate is below 1%; the dataset is released under CC BY 4.0 and the code under Apache-2.0 (GitHub). The corpus meets it once, as the benchmark on which OpenAI claimed GPT-5.6 Luna beats Claude Fable 5 at 99% lower cost per task in the July 2026 price cut.
Benchmark
Agents' Last Exam
Agent benchmark on which OpenAI claimed GPT-5.6 Luna beats Claude Fable 5 at 99% lower cost per task.
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to July 30, 2026 (theinnermostloop.substack.com)
Development timeline
Newest reporting dates first. Reporting dates may differ from event dates.
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.