Kernel Generation on KernelBench Level 3 (Hard)
98Correctness (Corr)Claude-4.6-opus
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Claude-4.6-opusFramework=Kernel-Smith evolution-agent, Evolution Rounds=402026.03 | 98 | 62 | 2.02 | |
| Kernel-Smith-235B-RLFramework=Kernel-Smith evolution-agent, Evolution Rounds=402026.03 | 94 | 46 | 1.02 | |
| DeepSeek-v3.2-SpecialeFramework=Kernel-Smith evolution-agent, Evolution Rounds=402026.03 | 90 | 32 | 1.14 | |
| Gemini-3.0-proFramework=Kernel-Smith evolution-agent, Evolution Rounds=402026.03 | 88 | 50 | 1.26 | |
| Qwen3-235B-2507-thinkFramework=Kernel-Smith evolution-agent, Evolution Rounds=402026.03 | 84 | 42 | 0.76 | |
| Qwen3.5-397B-thinkFramework=Kernel-Smith evolution-agent, Evolution Rounds=402026.03 | 74 | 38 | 0.61 | |
| Kimi-K2.5Framework=Kernel-Smith evolution-agent, Evolution Rounds=402026.03 | 74 | 38 | 0.61 | |
| Minimax-M2.5Framework=Kernel-Smith evolution-agent, Evolution Rounds=402026.03 | 56 | 30 | 0.36 |