All resources

Resources

Benchmarks

Evaluation reports will live here once the harness output is ready to publish with enough context to reproduce it.

Current technical material

What a report will include

Required

Harness

Commands, dataset shape, workload mix, version, and commit hash.

Required

Environment

Hardware, OS, filesystem, runtime mode, cache settings, and model/provider settings.

Required

Results

Raw output, summary charts, interpretation, and known limits.