Coding on MultiPL-E
87.9ScoreSOTA with Equivalent Parameters
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SOTA with Equivalent ParametersModel comparison=Equivalent Parameters2026.05 | 87.9 | — | |
| JT-Safe-V2-35BParameters=35B2026.05 | 82.47 | — | |
| LLaDA2.0-flash2026.02 | 74.87 | 314 | |
| LLaDA2.1-flashInference Mode=Q Mode2026.02 | 73.34 | 433 | |
| LLaDA2.1-flashInference Mode=S Mode2026.02 | 70.89 | 777 | |
| Qwen3-30B-A3B-Inst-25072026.02 | 70.67 | 100 | |
| Qwen2-72B-InstructType=Instruction-tuned2024.07 | 69.2 | — | |
| LLaDA2.0-mini2026.02 | 67.46 | 278 | |
| LLaDA2.1-minimode=Q Mode2026.02 | 67.17 | 401 | |
| Ling-mini-2.02026.02 | 67.09 | — | |
| Ling-flash-2.02026.02 | 65.76 | 100 | |
| LLaDA2.1-minimode=S Mode2026.02 | 64.16 | 723 | |
| Llama-3-70B-InstructType=Instruction-tuned2024.07 | 63.4 | — | |
| Qwen3-8Bno think=true2026.02 | 61.7 | — | |
| Mixtral-8x22B-InstructType=Instruction-tuned2024.07 | 61.1 | — | |
| Qwen1.5-110B-ChatType=Instruction-tuned2024.07 | 55.4 | — | |
| Qwen1.5-72B-ChatType=Instruction-tuned2024.07 | 48.1 | — | |
| DSRMidtraining Schedule=DSR2026.05 | 47.53 | — | |
| Qwen2-7BParameters=7B2024.07 | 46.3 | — | |
| WSDMidtraining Schedule=WSD2026.05 | 44.87 | — | |
| Cosine-decayMidtraining Schedule=Cos2026.05 | 43.95 | — | |
| Unified 15B MoEMidtraining Corpus=1k→8k2026.05 | 43.3 | — | |
| Unified 15B MoEMidtraining Corpus=4k→8k2026.05 | 42.54 | — | |
| Unified 15B MoEMidtraining Corpus=4k2026.05 | 41.11 | — | |
| Gemma-7BParameters=7B2024.07 | 29.7 | — | |
| Mistral-7BParameters=7B2024.07 | 29.4 | — | |
| Qwen1.5-7BParameters=7B2024.07 | 28.1 | — | |
| Llama-3-8BParameters=8B2024.07 | 22.6 | — | |
| Unified 15B MoEPretraining Corpus=4k→8k2026.05 | 19.74 | — | |
| Unified 15B MoEPretraining Corpus=1k→8k2026.05 | 18.34 | — | |
| Unified 15B MoEPretraining Corpus=4k2026.05 | 17.26 | — |