Optical Character Recognition on OCRBench
922ScoreRouting
Evaluation Results
| Method | Links | |||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RoutingEvaluation Strategy=Adaptive Routing, Aggregation Method=Best Model, Backbone=Qwen2.5VL-72B2025.04 | 922 | — | — | — | — | 8.2 | — | — | — | — | — | — | — | — | — | — | — | |
| RoutingEvaluation Strategy=Adaptive Routing, Aggregation Method=Average2025.04 | 907 | — | — | — | — | 6.5 | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-8BModel Category=Open-source General Models2025.11 | 906 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BasePruning Ratio=100%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 903 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VLModel Size=8B2026.03 | 896 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SingleEvaluation Strategy=Single model, Aggregation Method=Best Model, Backbone=Qwen2VL-72B2025.04 | 888 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-OV-7BModel Size=7B2026.02 | 882 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepEyesV2Tool=General, Param Size=7B2025.11 | 882 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Gemini 2.5 ProTool=Code, Param Size=-2025.11 | 881 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3-8BModel Size=8B2026.02 | 880 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3Tool=✗, Param Size=8B2025.11 | 880 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qianfan-OCR2026.03 | 880 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3-8BModel Category=Open-source General Models2025.11 | 880 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SC@3Evaluation Strategy=Self-Consistency (3 samples), Aggregation Method=Best Model, Backbone=Qwen2VL-72B2025.04 | 875 | — | — | — | — | 2.7 | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-4BParameters=4B2026.03 | 873 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeFacto (Ours)Backbone=Qwen2.5-VL-7B2025.09 | 871 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-2B# Total Params=2.1B, Evaluation Toolkit=VLMEvalKit2025.11 | 864 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| QwenVL2.5-7BModel Size=7B2026.02 | 864 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VLTool=✗, Param Size=7B2025.11 | 864 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ThymeTool=Code, Param Size=7B2025.11 | 863 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SenseNova-SIBase Architecture=Qwen3-VL-8B, Model Category=Ours2025.11 | 863 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VLParameters=2B2026.03 | 858 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Robobrain-2.0-7BModel Size=7B2026.02 | 857 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-2BParameters=2B2026.03 | 856 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VST-7B-SFTModel Category=Open-source SI Models2025.11 | 855 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-8BRetained Tokens=1024 Tokens (100%)2026.07 | 855 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3-2BModel Category=Open-source General Models2025.11 | 853 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Penguin-VLModel Size=8B2026.03 | 852 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SingleEvaluation Strategy=Single model, Aggregation Method=Average2025.04 | 852 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FP16Model=MiniCPM-V-2.6-8B, Spiking=false, Time Step=N/A2026.04 | 850 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL-3.5Model Size=8B2026.03 | 840 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3.5Parameters=2B2026.03 | 836 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3-2BModel Size=2B2026.02 | 835 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FP16Model=Qwen2VL-7B, Spiking=false, Time Step=N/A2026.04 | 835 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3.5-2B# Total Params=2.3B, Evaluation Toolkit=VLMEvalKit2025.11 | 834 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RoboInter-Qwen-7BModel Size=7B, Backbone=Qwen2026.02 | 832 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(QuaRot+MSTS+TC-LIF)Model=MiniCPM-V-2.6-8B, Spiking=true, Time Step=3/42026.04 | 829 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=100%2026.05 | 829 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| NGMModel=Qwen3-VL-2B2026.05 | 829 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SC@3Evaluation Strategy=Self-Consistency (3 samples), Aggregation Method=Average2025.04 | 828 | — | — | — | — | -2.8 | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(QuaRot)Model=MiniCPM-V-2.6-8B, Spiking=true, Time Step=152026.04 | 828 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Q2.5VL-7BTokens Retained Percentage=100%2025.03 | 827 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(GPTQ)Model=MiniCPM-V-2.6-8B, Spiking=true, Time Step=2552026.04 | 827 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen3-VL-2BModel=Qwen3-VL-2B2026.05 | 826 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TwigVLM++Tokens Retained Percentage=33.3%2025.03 | 825 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-3B# Total Params=3.75B, Evaluation Toolkit=VLMEvalKit2025.11 | 824 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(QuaRot+MSTS+TC-LIF)Model=Qwen2VL-7B, Spiking=true, Time Step=3/42026.04 | 823 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LFM2-VL-3B# Total Params=3.0B, Evaluation Toolkit=VLMEvalKit2025.11 | 822 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4oTool=Code, Param Size=-2025.11 | 822 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VisionZipBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=70%2026.05 | 819 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SenseNova-SIBase Architecture=InternVL3-2B, Model Category=Ours2025.11 | 817 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LaVerVisual Encoder Model=Qwen-ViT2025.12 | 815 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SenseNova-SIBase Architecture=InternVL3-8B, Model Category=Ours2025.11 | 815 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaselineVisual Encoder Model=Qwen-ViT2025.12 | 813 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TwigVLM++Tokens Retained Percentage=22.2%2025.03 | 813 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Robobrain-2.0-3BModel Size=3B2026.02 | 811 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Bagel-7B-MoTModel Category=Open-source General Models2025.11 | 811 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Penguin-VLParameters=2B2026.03 | 810 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(RTN)Model=MiniCPM-V-2.6-8B, Spiking=true, Time Step=2552026.04 | 810 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VanillaBackbone=DeepSeek-VL2, Pruning Ratio=100%2026.05 | 803 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VisMMOEBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=70%2026.05 | 801 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ERASEPruning Ratio=-65.11%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 800 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3-1BModel Size=1B2026.02 | 798 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| QwenVL2.5-3BModel Size=3B2026.02 | 797 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SenseNova-SIBase Architecture=Bagel-7B-MoT, Model Category=Ours2025.11 | 797 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| FP16Model=InternVL2-8B, Spiking=false, Time Step=N/A2026.04 | 795 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MixBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=70%2026.05 | 793 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(QuaRot)Model=Qwen2VL-7B, Spiking=true, Time Step=152026.04 | 792 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(QuaRot+MSTS+TC-LIF)Model=InternVL2-8B, Spiking=true, Time Step=3/42026.04 | 791 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternVL3.5-1B# Total Params=1.1B2025.11 | 790 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RoboInter-Qwen-3BModel Size=3B, Backbone=Qwen2026.02 | 787 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VisionZipBackbone=DeepSeek-VL2, Pruning Ratio=70%2026.05 | 780 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TwigVLM++Tokens Retained Percentage=11.1%2025.03 | 772 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(QuaRot)Model=InternVL2-8B, Spiking=true, Time Step=152026.04 | 772 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VisMMOEBackbone=DeepSeek-VL2, Pruning Ratio=70%2026.05 | 771 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| IVC-PrunePruning Ratio=-65.11%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 766 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MixBackbone=DeepSeek-VL2, Pruning Ratio=70%2026.05 | 761 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ERASEPruning Ratio=-72.77%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 761 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(GPTQ)Model=Qwen2VL-7B, Spiking=true, Time Step=2552026.04 | 760 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternViT-2.5-6BBase LLM=Qwen2.5-7B, Size=6.0B, AnyRes=false, Discrete=false2026.06 | 757 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| InternViT-2.5-6BBase LLM=Qwen2.5 - 7B, Size=6.0B, AnyRes=false, Discrete=false2026.06 | 757 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Gemini-1.5-Pro2025.09 | 754 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| iLLaVAPruning Ratio=-65.11%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 752 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-VIFParams=7B, Base model=Qwen-2.5-VL2026.05 | 747 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| TwigVLMTokens Retained Percentage=33.3%2025.03 | 744 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| VisionZipBackbone=Qwen3-VL-30B-A3B, Pruning Ratio=50%2026.05 | 741 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(QuaRot+MSTS+TC-LIF)Model=MiniCPM-V-2.6-8B, Spiking=true, Time Step=2/32026.04 | 739 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4o2025.09 | 736 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VLParams=7B, Base model=Qwen-2.5-VL2026.05 | 732 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LFM2-VL-1.6B# Total Params=1.6B2025.11 | 729 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SmolVLM2Parameters=2.2B2026.03 | 729 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PruneSIDPruning Ratio=-65.11%, Base Model=Qwen3-VL-8B, Zero-shot evaluation=True2026.05 | 726 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SmolVLM2-2.2B# Total Params=2.2B, Evaluation Toolkit=VLMEvalKit2025.11 | 725 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RoboInter-LLaVAOV-7BModel Size=7B, Backbone=LLaVA-OV2026.02 | 725 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-SFTParams=7B, Base model=Qwen-2.5-VL2026.05 | 724 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| SpikeMLLM(QuaRot+MSTS+TC-LIF)Model=Qwen2VL-7B, Spiking=true, Time Step=2/32026.04 | 717 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| X-Omni2026.03 | 712 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViQBase LLM=Qwen2.5-7B, Size=1.3B, AnyRes=true, Discrete=true2026.06 | 711 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ViQBase LLM=Qwen2.5 - 7B, Size=1.3B, AnyRes=true, Discrete=true2026.06 | 711 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GvU2026.03 | 709 | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — | — |