Skip to content

Benchmark

Kernel Bench L3 · Win Rate

Scores are percentages. Higher is better.

Models measured
2
Scores recorded
2
Sources
1
IQ weight
No weight published

Score distribution

97%
Mean
98%
Median
1%
Std dev
96%
10th percentile
98%
90th percentile
96%
Min
96%98%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1Qwen3.7 PlusQwen3.7-Plus release · 50 problems · isolated H100 80GB container · CUTLASS/CUDA docs only · 500 tool-call cap · fraction faster than torch.compile· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.98%
2Qwen3.7 MaxQwen3.7-Max release · 50 problems · isolated H100 80GB container · CUTLASS/CUDA docs only · 500 tool-call cap · fraction faster than torch.compile· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.96%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.