Document Question Answering on DocVQA (test)
96.4AccuracyQwen2.5-VL-72B
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-VL-72BModel Size=72B2025.02 | 96.4 | |
| Qwen3-VL-8BAccess=Open weights only2026.03 | 96.1 | |
| Qwen2.5-VL-7BModel Size=7B2025.02 | 95.7 | |
| Qwen2.5-VL-7BJudge=GPT-4o2025.12 | 95.5 | |
| MiMo-VL-7B-SFT-2508Judge=GPT-4o2025.12 | 95.5 | |
| Qwen3-VL-4BAccess=Open weights only2026.03 | 95.3 | |
| Claude 3.52024.07 | 95.2 | |
| Claude-3.5 Sonnet2025.02 | 95.2 | |
| MiMo-VL-Miloco-7BJudge=GPT-4o2025.12 | 95.2 | |
| InternVL2.5-78BModel Size=78B2025.02 | 95.1 | |
| MiniCPM-V-4.5-8BAccess=Open weights only2026.03 | 94.7 | |
| PLM-8BAccess=Fully open2026.03 | 94.6 | |
| Qwen2-VLSize=7B, #token /image tile=Native resolution (unfixed)2024.12 | 94.5 | |
| Qwen2-VL# Parameters=8B2024.12 | 94.5 | |
| Eagle2.5-8BAccess=Open weights only2026.03 | 94.1 | |
| NVILA# Parameters=15B2024.12 | 94 | |
| Qwen2.5-VL-3BModel Size=3B2025.02 | 93.9 | |
| InternVL2# Parameters=40B2024.12 | 93.9 | |
| PLM-3BAccess=Fully open2026.03 | 93.8 | |
| MolmoPoint-8BAccess=MolmoPoint2026.03 | 93.8 | |
| NVILA# Parameters=8B2024.12 | 93.7 | |
| Keye-VL-1.5-8BAccess=Open weights only2026.03 | 93.4 | |
| GLM-4.1V-9BAccess=Open weights only2026.03 | 93.3 | |
| Molmo2-8BAccess=Fully open2026.03 | 93.2 | |
| Gemini 1.5 Proexternal_ocr_tools=true2024.07 | 93.1 | |
| Gemini-1.5-Pro2024.12 | 93.1 | |
| Gemini 1.5 Pro2025.02 | 93.1 | |
| Gemini 1.5 Pro2024.12 | 93.1 | |
| GPT-4o2024.07 | 92.8 | |
| GPT-4o2024.12 | 92.8 | |
| GPT-4o2024.12 | 92.8 | |
| Llama 3-V 405B2024.07 | 92.6 | |
| NVLM-D-1.0# Parameters=78B2024.12 | 92.6 | |
| PVC InternVL2 (Ours)Size=8B, #token /image tile=642024.12 | 92.5 | |
| InternVL3.5-4BAccess=Open weights only2026.03 | 92.4 | |
| InternVL3.5-8BAccess=Open weights only2026.03 | 92.3 | |
| Llama 3-V 70B2024.07 | 92.2 | |
| NVILA-Lite# Parameters=8B2024.12 | 91.7 | |
| Claude Sonnet 4.5Access=API call only2026.03 | 91.7 | |
| InternVL2Size=8B, #token /image tile=2562024.12 | 91.6 | |
| InternVL2# Parameters=8B2024.12 | 91.6 | |
| Gemini 2.5 ProAccess=API call only2026.03 | 91.5 | |
| LLaVA-OneVision# Parameters=72B2024.12 | 91.3 | |
| GPT-4o2025.02 | 91.1 | |
| Gemini 2.5 FlashAccess=API call only2026.03 | 91.1 | |
| IXC-2.5Size=7B, #token /image tile=4002024.12 | 90.9 | |
| Pixtral# Parameters=12B2024.12 | 90.7 | |
| NVILA-Lite# Parameters=15B2024.12 | 90.6 | |
| MolmoPoint-8B-O-7BAccess=Fully open2026.03 | 90.4 | |
| Qwen2-VLSize=2B, #token /image tile=Native resolution (unfixed)2024.12 | 90.1 | |
| Llama 3.2# Parameters=90B2024.12 | 90.1 | |
| OracleModel=MiniCPM-V 2.62026.02 | 89.41 | |
| InternVL3-8BJudge=GPT-4o2025.12 | 89.4 | |
| Oryx MLLMSize=7B2024.12 | 89 | |
| GPT-5Access=API call only2026.03 | 88.9 | |
| GPT-4V2024.07 | 88.4 | |
| GPT-4V2024.12 | 88.4 | |
| Llama 3.2# Parameters=11B2024.12 | 88.4 | |
| GPT-4VModel Category=Closed-source Models2025.12 | 88.4 | |
| SDAR-VL-8B-InstModel Category=Block-wise Discrete Diffusion Models2025.12 | 88.3 | |
| OracleModel=GPT-4o2026.02 | 88.1 | |
| Molmo2-4BAccess=Fully open2026.03 | 87.8 | |
| PVC InternVL2 (Ours)Size=2B, #token /image tile=642024.12 | 87.7 | |
| LLaVA-OVSize=7B, #token /image tile=7292024.12 | 87.5 | |
| LLaVA-OneVision# Parameters=8B2024.12 | 87.5 | |
| InternVL2Size=2B, #token /image tile=2562024.12 | 86.9 | |
| GPT-5 miniAccess=API call only2026.03 | 86.7 | |
| Gemma-3-27B-ITJudge=GPT-4o2025.12 | 86.6 | |
| Claude 3.5 Sonnet2024.12 | 85.2 | |
| Florence-VL# Parameters=8B2024.12 | 84.9 | |
| MARAModel=GPT-4o2026.02 | 84.64 | |
| Llama 3-V 8B2024.07 | 84.4 | |
| MARA (Our Framework)Model=GPT-4o, Input=top-32026.02 | 81.22 | |
| LongVASize=7B, #token /image tile=1442024.12 | 80.8 | |
| MARAModel=MiniCPM-V 2.62026.02 | 80.59 | |
| VisRAGModel=GPT-4o, Input=top-32026.02 | 78.21 | |
| Cambrian-1# Parameters=8B2024.12 | 77.8 | |
| Cambrian-1# Parameters=13B2024.12 | 76.8 | |
| MARA (Our Framework)Model=MiniCPM-V 2.6, Input=top-32026.02 | 76.53 | |
| Cambrian-1# Parameters=34B2024.12 | 75.5 | |
| VisRAGModel=MiniCPM-V 2.6, Input=top-32026.02 | 74.5 | |
| MARA (Our Framework)Model=GPT-4o, Input=top-12026.02 | 70.28 | |
| MARA (Our Framework)Model=MiniCPM-V 2.6, Input=top-12026.02 | 69.45 | |
| VisRAGModel=MiniCPM-V 2.6, Input=top-12026.02 | 69.4 | |
| VisRAGModel=GPT-4o, Input=top-12026.02 | 68.32 | |
| MARA (Our Framework)Model=MiniCPM-V 2.6, Input=top-102026.02 | 60.56 | |
| VisRAGModel=MiniCPM-V 2.6, Input=top-102026.02 | 59.07 | |
| VILA-1.5# Parameters=13B2024.12 | 58.6 | |
| VILA-1.5# Parameters=40B2024.12 | 58.6 | |
| VILA-1.5# Parameters=8B2024.12 | 40.6 | |
| LLaVA-1.5# Parameters=13B2024.12 | 30.3 | |
| LLaVA-1.5# Parameters=7B2024.12 | 28.1 |