Benchmark
KernelBench Pass Rate
Coding benchmarks compare models on programming tasks. Scores are percentages. Higher is better.
All coding benchmarksScores from KernelBench (kernelbench.com) data.
- Models measured
- 15
- Scores recorded
- 15
- Sources
- 1
- IQ weight
- 0.00
Score distribution
- 78%
- Mean
- 83.3%
- Median
- 16.3%
- Std dev
- 60%
- 10th percentile
- 100%
- 90th percentile
- 50%
- Min
This benchmark has a weight of 0.00 for chat models, measured over 15 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5hard· KernelBench (kernelbench.com) data | 100% |
| 2 | DeepSeek Flashhard· KernelBench (kernelbench.com) data | 100% |
| 3 | DeepSeek V4 Prohard· KernelBench (kernelbench.com) data | 100% |
| 4 | DeepSeek V4.1 Flash 1017hard· KernelBench (kernelbench.com) data | 100% |
| 5 | Claude Fable 5hard· KernelBench (kernelbench.com) data | 83.3% |
| 6 | GLM 5.2hard· KernelBench (kernelbench.com) data | 83.3% |
| 7 | Grok 4.6hard· KernelBench (kernelbench.com) data | 83.3% |
| 8 | LongCat 2.0hard· KernelBench (kernelbench.com) data | 83.3% |
| 9 | Claude Opus 4.8hard· KernelBench (kernelbench.com) data | 66.7% |
| 10 | DeepSeek V4 Flash 0731hard· KernelBench (kernelbench.com) data | 66.7% |
| 11 | Grok 4.5hard· KernelBench (kernelbench.com) data | 66.7% |
| 12 | MiniMax M3hard· KernelBench (kernelbench.com) data | 66.7% |
| 13 | GLM 5.3hard· KernelBench (kernelbench.com) data | 60% |
| 14 | Qwen3.8 Maxhard· KernelBench (kernelbench.com) data | 60% |
| 15 | GPT-5.6 Solhard· KernelBench (kernelbench.com) data | 50% |