Demo

Demo desk. Every organisation, source and story here is fictional and frozen at the demo clock. Nothing on this site is current reporting. Demo clock: Thu 12 Mar 2026, 09:00 UTC, frozen. About the demo

Evaluation

Benchmarks, leaderboards, test contamination and how results are measured.

2 events · heat 64.0 at the demo clock · +47.9 in 24 h

33.9

New Open weights · Model releases Last report 12 Mar, 07:30

Introducing Orchard 3: open-weight models in three sizes

5 independent sources · 6 reports · 1 syndicated copy not counted · 1 duplicate dropped
30.1

Follow-up Evaluation Last report 12 Mar, 07:00

ReasonBench v2: test items found in public training data; scores under review

Latest ReasonBench corrections reshuffle the top of the leaderboard

5 independent sources · 6 reports

Other topics