General Reasoning on General benchmark Hard
55.8MMLU AccuracyHydraHead
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| HydraHeadType=Hybrid2026.06 | 55.8 | 46.64 | 47.2 | 52.84 | 50.62 | |
| Qwen3-1.7BType=FA2026.06 | 55.46 | 48.16 | 43 | 69.45 | 54.02 | |
| Gemma-3n-E2BType=Hybrid2026.06 | 53.82 | 42.47 | 35 | 25.85 | 39.29 | |
| Jet-Nemotron-2BType=Hybrid2026.06 | 53.59 | 58.16 | 54.2 | 75.28 | 60.31 | |
| Qwen3-1.7B-yarnType=FA2026.06 | 53.45 | 45.65 | 42.4 | 59.97 | 50.37 | |
| Hymba-1.5BType=Hybrid2026.06 | 49.67 | 8.89 | 16.6 | 0.994 | 19.04 | |
| HypeNet-2BType=Hybrid2026.06 | 46.52 | 9.15 | 4.8 | 4.93 | 16.35 | |
| MiniCPM5-1B-BaseType=FA2026.06 | 39.3 | 37.81 | 33.2 | 34.12 | 36.11 |