Hierarchical Task Decomposition on MagicEval-Plan Condition 3
97.5Step CountMagicAgent-32B
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MagicAgent-32BScale=MagicAgent Series2026.02 | 97.5 | 95.3 | 87.4 | |
| MagicAgent-30B-A3BScale=MagicAgent Series2026.02 | 97 | 95.5 | 88.1 | |
| Kimi-K2-InstructScale=Ultra-Scale Models2026.02 | 87 | 91 | 80.2 | |
| GLM-4.7Scale=Ultra-Scale Models2026.02 | 85.5 | 96.6 | 83.2 | |
| Qwen3-MAXScale=Ultra-Scale Models2026.02 | 76 | 91 | 80.6 | |
| Qwen3-235B-A22B-InstructScale=Ultra-Scale Models2026.02 | 72.5 | 89.6 | 74.6 | |
| Qwen3-32B-nothinkScale=Large-Scale Models2026.02 | 67.5 | 90.4 | 77.1 | |
| DeepSeek-V3.1-nothinkScale=Ultra-Scale Models2026.02 | 65 | 96.5 | 77 | |
| GPT-5.2Scale=Ultra-Scale Models2026.02 | 51 | 89.9 | 85.3 | |
| ERNIE-4.5-21B-A3B-PTScale=Large-Scale Models2026.02 | 49 | 90.2 | 76.2 | |
| Qwen3-30B-A3B-InstructScale=Large-Scale Models2026.02 | 46.5 | 89.8 | 74.4 | |
| Llama3.3-70B-InstructScale=Large-Scale Models2026.02 | 40.5 | 87.9 | 70.2 | |
| Olmo-3.1-32B-InstructScale=Large-Scale Models2026.02 | 39.5 | 87.7 | 71.9 | |
| GLM-4.7-FlashScale=Large-Scale Models2026.02 | 37 | 83.8 | 65.6 |