BENCHMARK OBSERVATORY

Leaderboard of Benchmarks

Models are probes. The benchmark is the object we compare.

BLE evaluates benchmarks themselves, not where their results are reported. It is deliberately selective: a small set of classic, widely followed benchmarks, chosen to show frontier progress with relatively little information.