Code Generation on HumanEval+ (HEVAL+)
73.8HEVAL+ ScoreQWEN-3-30B MOE
Evaluation Results
| Method | Links | |
|---|---|---|
| QWEN-3-30B MOEPARAMS=30B, Variant=ORIGINAL2025.11 | 73.8 | |
| QWEN-3-30B MOEPARAMS=30B, Variant=FCSD2025.11 | 73.2 | |
| LLAMA-4-SCOUTPARAMS=109B, Variant=FCSD2025.11 | 63.4 | |
| LLAMA-4-SCOUTPARAMS=109B, Variant=ORIGINAL2025.11 | 62.2 | |
| DEEPSEEK-V2-LITEPARAMS=16B, Variant=FCSD2025.11 | 41.5 | |
| DEEPSEEK-V2-LITEPARAMS=16B, Variant=ORIGINAL2025.11 | 40.2 | |
| LLAMA-4-SCOUTPARAMS=109B, Variant=SFT2025.11 | 14 | |
| DEEPSEEK-V2-LITEPARAMS=16B, Variant=SFT2025.11 | 11 | |
| QWEN-3-30B MOEPARAMS=30B, Variant=SFT2025.11 | 0.6 |