Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Kernel Generation on KernelBench Level 3 (Hard)
Loading...
98
Correctness (Corr)
Claude-4.6-opus
54.32
65.66
77
88.34
Mar 30, 2026
Correctness (Corr)
Fast1 Score
Average AMSR
Updated 19d ago
Evaluation Results
Method
Method
Links
Correctness (Corr)
Fast1 Score
Average AMSR
Claude-4.6-opus
Framework=Kernel-Smith...
2026.03
98
62
2.02
Kernel-Smith-235B-RL
Framework=Kernel-Smith...
2026.03
94
46
1.02
DeepSeek-v3.2-Speciale
Framework=Kernel-Smith...
2026.03
90
32
1.14
Gemini-3.0-pro
Framework=Kernel-Smith...
2026.03
88
50
1.26
Qwen3-235B-2507-think
Framework=Kernel-Smith...
2026.03
84
42
0.76
Qwen3.5-397B-think
Framework=Kernel-Smith...
2026.03
74
38
0.61
Kimi-K2.5
Framework=Kernel-Smith...
2026.03
74
38
0.61
Minimax-M2.5
Framework=Kernel-Smith...
2026.03
56
30
0.36
Feedback
Search any
task
Search any
task