The Aggregate (theaggregate.ai)
Website by The Aggregate
Unified rankings for 10514 ranked models (2287 in the fused corpus) across 35952 public benchmarks, of which 29208 are listed in the current ranking. Refreshed daily. A predictive-evidence model combines measured results using benchmark predictive error and calibration uncertainty. Targets have no minimum benchmark or source count; only connected evidence contributes to a rating. The displayed ± is conditional capability uncertainty. Rows below are fused across reasoning-effort settings, one per model, matching the default board on /models. Human reference baselines combine benchmark-specific human results and estimates and are excluded from the ranking. Page HTML served to crawlers contains readable summaries of every model and benchmark.
Verification: observed. Source: crawl.