Skip to content

Benchmark

KernelBench Hard

Coding benchmarks compare models on programming tasks. Scores are percentages. Higher is better.

Models measured
1
Scores recorded
3
Sources
2
IQ weight
No weight published

Score distribution

28.8%
Mean
28.8%
Median
0%
Std dev
28.8%
10th percentile
28.8%
90th percentile
28.8%
Min

Leaderboard

#ModelScore
1MiniMax M3Claude Code · NVIDIA Blackwell sm_120 · 9-task mean· Reported by MiniMax Research· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.28.8%
2MiniMax M3official model-page overview figure· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.28.8%
3MiniMax M3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim28.8%

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. · Data by LLM Stats