Benchmark
Filter
Scores in 10-point tiers.
Tier S starts at the best score, rounded to 5. Up to 60 models shown.
open
Every reported score.
Best published score per model. Fu does not re-run evaluations.
| # | Model | Score | Coverage | Generation | Model family | Lab | Params | Released |
|---|
Dashed line: least-squares fit. Rules: best open (green) and closed (pink) score.