Evidence before rankings
Proof Lab benchmarks.
These are published test records, not a mysterious universal score. Open the method, result summary, tested version and evidence link before treating a result as relevant to your own work.
No published benchmark results match this view yet. ToolOrbitHub will not fill the gap with unsupported rankings.
Want a test added?
A useful benchmark needs a repeatable task, a stated method and an evidence trail. Submit a correction or firsthand experience from a tool profile, or use the Control Center Proof Lab when you are ready to publish a supported test.
Browse AI categories