Multimodal Reasoning on OckBench 20-eval suite (test)
71.9AccuracyQwen3-VL-4B-Instruct
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Qwen3-VL-4B-InstructN (B)=4.32026.06 | 71.9 | 1.3 | 108.9 | |
| Qwen3.5-4BN (B)=4.02026.06 | 70.4 | 14.58 | 1,284 | |
| Qwen3.5-2BN (B)=2.02026.06 | 69.5 | 0.97 | 169.4 | |
| Datology 4BN (B)=4.42026.06 | 69.1 | 0.41 | 32.5 | |
| Qwen3-VL-2B-InstructN (B)=2.12026.06 | 67.4 | 0.63 | 101.6 | |
| Datology 2BN (B)=2.12026.06 | 66 | 0.27 | 42.1 | |
| InternVL3.5-4BN (B)=4.32026.06 | 65.8 | 1.12 | 85.4 | |
| Datology 1BN (B)=1.02026.06 | 63.8 | 0.1 | 32.4 | |
| Perceptron-Isaac-2BN (B)=2.12026.06 | 63.1 | 0.6 | 89.8 | |
| InternVL3-2BN (B)=2.12026.06 | 58.9 | 0.21 | 30 | |
| InternVL3.5-2BN (B)=2.32026.06 | 56 | 1.06 | 128.6 | |
| Mammoth 4BN (B)=4.42026.06 | 54.9 | 0.96 | 59.6 | |
| Mammoth 2BN (B)=2.12026.06 | 54.1 | 0.43 | 55.8 | |
| Mammoth 1BN (B)=1.02026.06 | 50.5 | 0.23 | 57.3 |