Fu Ranking
1,493 models ranked across 713 benchmarks, adjusted for how hard each benchmark is.
Fu Score: expected % on a benchmark of median difficulty. Updated 2026-10-24.
scores39,407
benchmarks713
pairs ordered right, Fu84.7%
pairs ordered right, plain avg67.4%
All models
Bar: 90% rank interval (range of ranks when the benchmark set is resampled), log axis.
| Rank | Model | Fu score | 90% rank interval log | Benchmarks | Avg. rank |
|---|
No models match. Clear the search or untick “Hide uncertain ranks”.
90% rank interval Fu rank plain-average rank
⚑ Selective reporting: rank drops if skipped popular benchmarks are assumed to have gone badly. Hidden models keep their place in the ranks.
Average rank vs. Fu rank.
Top 15: plain-average rank (left) vs. Fu rank (right). Highlighted: the three biggest moves.
plain average → Fu rankrank, log scale, top 15━ 3 biggest moves
Compare two models.
90% rank interval━ Model A ━ Model B ◆ selective-reporting rank
Expected score on well-known benchmarks
Top models per task universe
Each universe is ranked on its own benchmarks only.