Code Generation on MBPP CodaSet OOD (test)
79.37Performance (%)Qwen3-235B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-235B2026.05 | 79.37 | 13.8 | |
| Kimi-K2-Instruct2026.05 | 77.78 | 20.1 | |
| DecoRBase Model=DeepSeek-V3.12026.05 | 76.72 | 12.4 | |
| gemma-3-27b-it2026.05 | 76.19 | 1.3 | |
| gemma-3-12b-it2026.05 | 76.19 | 1 | |
| DeepSeek-V3.12026.05 | 75.66 | 14.1 | |
| DeepSeek-V3.2-Exp2026.05 | 75.13 | 12.2 | |
| MODEL-SAT2026.05 | 74.6 | 4.7 | |
| EmbedLLM2026.05 | 73.72 | 8.4 | |
| gpt-oss-120b2026.05 | 73.54 | 11.6 | |
| KNN Router2026.05 | 73.54 | 2.7 | |
| LLM Router2026.05 | 73.19 | 4 | |
| Mistral-3.2-24B2026.05 | 73.02 | 6 | |
| RouterDC2026.05 | 72.66 | 6.9 | |
| DecoRBase Model=gpt-oss-120b2026.05 | 72.49 | 10.2 | |
| Random Router2026.05 | 71.96 | 8.9 |