OCR-related understanding on AI2D (test)
85AccuracyGLM-4.6V-FlashX
Evaluation results
| Method | Links | |
|---|---|---|
| GLM-4.6V-FlashX2026.01 | 85 | |
| Ovis2.5Parameter Scale=9B2026.01 | 84.6 | |
| Qwen3-VLParameter Scale=8B2026.01 | 84.5 | |
| Qwen3-VLModel Size=4B, Training Stage=instruct2026.05 | 84.1 | |
| Ostrakon-VLParameter Scale=8B2026.01 | 83 | |
| InternVL-3.5Model Size=4B2026.05 | 82.6 | |
| FineViT-VLLanguage Backbone=Qwen3 1.7B LLM, Evaluation Framework=VLMEvalKit [69]2026.03 | 82.25 | |
| Qwen3-VLModel Size=4B, Training Stage=S-12026.05 | 81.6 | |
| Qwen2.5-VLParameter Scale=72B2026.01 | 80.6 | |
| InternVL3.5Parameter Scale=8B2026.01 | 79.7 | |
| Qwen3-VL-SegModel Size=4B, Training Stage=S-22026.05 | 79.7 | |
| Intern3.5-VLLanguage Backbone=Qwen3 1.7B LLM, Evaluation Framework=VLMEvalKit [69]2026.03 | 79.01 | |
| Qwen3-VLLanguage Backbone=Qwen3 1.7B LLM, Evaluation Framework=VLMEvalKit [69]2026.03 | 76.46 | |
| Aquila-VLLanguage Backbone=Qwen2.5 1.5B LLM, Evaluation Framework=VLMEvalKit [69]2026.03 | 75.06 |