Holistic Evaluation on Combined Suite General Reasoning Perception Text
76.3Text AverageQwen3-32B
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Qwen3-32BTarget LLM=Public Models2026.05 | 76.3 | — | — | — | — | — | |
| Qwen3-4BTarget LLM=Public Models2026.05 | 67.9 | — | — | — | — | — | |
| DPA-Qwen3-32BTarget LLM=Qwen3-32B2026.05 | 58.1 | 60.9 | 50.1 | 77.9 | 61.1 | 60.3 | |
| LLaVA-NeXT-Qwen3-32BTarget LLM=Qwen3-32B2026.05 | 53.1 | 57.6 | 48.3 | 73.4 | 58.1 | 56.7 | |
| DPA-Qwen3-4BTarget LLM=Qwen3-4B2026.05 | 52.6 | 52.5 | 41 | 72.4 | 53.1 | 53 | |
| LLaVA-NeXT-Qwen3-4BTarget LLM=Qwen3-4B2026.05 | 45.1 | 51.1 | 40.1 | 68.3 | 51.2 | 49.6 | |
| Qwen2-VL-2BTarget LLM=Public Models2026.05 | 26.7 | 53.3 | 32.2 | 77.8 | 51.5 | 44.7 | |
| Idefics2-8BTarget LLM=Public Models2026.05 | 26.3 | 45.9 | 36.3 | 67.8 | 47.8 | 41.9 | |
| DPA-LLaMA-3.2-3BTarget LLM=LLaMA-3.2-3B2026.05 | 25.1 | 44.8 | 29.8 | 64.3 | 44.1 | 38.9 | |
| LLaVA-NeXT-LLaMA-3.2-3BTarget LLM=LLaMA-3.2-3B2026.05 | 21 | 40.8 | 27.4 | 60.5 | 40.7 | 35.3 | |
| LLaVA-1.5-7BTarget LLM=Public Models2026.05 | 19.3 | 34.9 | 23.7 | 36.5 | 31.1 | 27.9 | |
| LLaVA-NeXT-7BTarget LLM=Public Models2026.05 | 19.1 | 42.4 | 27.4 | 58.4 | 40.8 | 34.9 | |
| Cambrian-1-8BTarget LLM=Public Models2026.05 | 2.8 | 51.8 | 36.7 | 67.4 | 50 | 37.1 |