Visual Document Understanding on MMLongBenchDoc
58.3AccuracySynthetic Reasoning Qwen
Evaluation results
| Method | Links | |
|---|---|---|
| Synthetic Reasoning QwenParams=32B2026.03 | 58.3 | |
| Qwen3 VL 235B A22B InstructParams=235B (22B)2026.03 | 57 | |
| Qwen3 VL 235B A22B ThinkingParams=235B (22B)2026.03 | 56.2 | |
| TeleMM-2.02026.03 | 56.1 | |
| Qwen3 VL 32B InstructParams=32B2026.03 | 55.4 | |
| GLM 4.6VParams=106B (12B)2026.03 | 54.9 | |
| GPT-4o2026.03 | 46.3 |