Visual Document Understanding on VA
98.4AccuracyQwen3 VL 235B A22B Instruct
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3 VL 235B A22B InstructModel family=Qwen3 VL, Model scale=235B A22B, Reasoning strategy=Instruct2026.03 | 98.4 | |
| Synthetic ReasoningModel family=Qwen3 VL, Model scale=32B, Reasoning strategy=Synthetic Reasoning2026.03 | 95 | |
| LongPOModel family=Qwen3 VL, Reasoning strategy=Long-context Preference Optimization2026.03 | 94 | |
| Qwen3 VL 32B InstructModel family=Qwen3 VL, Model scale=32B, Reasoning strategy=Instruct2026.03 | 93.7 | |
| Plain DistillationModel family=Qwen3 VL, Reasoning strategy=Standard Distillation2026.03 | 92 | |
| No-thinkModel family=Qwen3 VL, Reasoning strategy=Direct Answer2026.03 | 91.7 | |
| Synthetic ReasoningModel family=Mistral, Model scale=24B, Reasoning strategy=Synthetic Reasoning2026.03 | 87.8 | |
| Plain DistillationModel family=Mistral, Reasoning strategy=Standard Distillation2026.03 | 84.4 | |
| Qwen Thinking TracesModel family=Mistral, Reasoning strategy=Thinking Traces2026.03 | 82.6 | |
| Mistral 3.1 Small 24BModel family=Mistral, Model scale=24B2026.03 | 80.2 | |
| No-thinkModel family=Mistral, Reasoning strategy=Direct Answer2026.03 | 77.9 |