Coming soon
Benchmarks
Every score on this page will come from a controlled CHAP evaluation of a pinned commit — nothing self-reported. No benchmarks have been run yet, and exactly how they will be run is still being decided, so results are not published yet.
What will appear here
One ranked table per reference dataset, with a full run record behind every row. The exact datasets, metrics and run parameters will be documented once the methodology is settled. Results land in the repo's benchmarks/ directory by pull request, like everything else.
Per-model benchmark charts get the same treatment — they appear on each model's detail page together with the first recorded runs.Browse models