The benchmark designed to outlast models is becoming their baseline. claude fable 5.1 leads…
the benchmark designed to outlast models is becoming their baseline. claude fable 5.1 leads humanity's last exam at 65% in the september update, ahead of opus 5 and mythos 5 across 62 models.
the exam that was meant to stay hard is now just another leaderboard.
Context
Humanity's Last Exam is a 2,500-question benchmark from CAIS and Scale. The Scale Labs leaderboard says CAIS released the September 17, 2026 version, HLE-Rolling, that the legacy version is called HLE-preview, and that all current model performance on this version is similar to the previous version. Two third-party aggregators, read at later dates, show Claude Fable 5.1 at 65%, Claude Opus 5 at 64.7% and a Mythos model at 64.5% to 64.7%. BenchLM, updated October 2, 2026, lists 63 models, and LLMBoard, updated October 1, 2026, lists 104.
The 65% and the ordering hold on those two aggregators only. The note's 62 models matches neither count. Other leaderboards disagree. Artificial Analysis lists Claude Opus 5.5 first at 61.4% in a listing dated October 3, 2026, and one other site lists Fable 5.1 at 59.1%, so scores from different leaderboards, harnesses and dates must not be merged. The Mythos model's name and score also differ between sources. The Scale page supports the benchmark version and not these scores. That the benchmark is becoming a baseline is the author's take, since top scores of about 65% on a 100% scale do not show saturation.
Watch next
- Scale and CAIS leaderboard values for the named models, and the number of models in the September update.
Sources
- Scale Labs: Humanity's Last Exam leaderboardlabs.scale.com
- BenchLM: HLEbenchlm.ai
- LLMBoard: Humanity's Last Examllmboard.ai
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 22 September 2026 at 02:04 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →