List your company
ResearchFormerly WizWand

Infographic Question Answering on InfoVQA

89.2ANLSQwen3-VL
23.26440.38257.574.618Oct 8, 2024Jan 9, 2025Apr 13, 2025Jul 15, 2025Oct 17, 2025Jan 18, 2026Apr 22, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen3-VLSize=235B(A22B)2026.0389.2
Qwen2.5-VLSize=72B2026.0387.3
InternVL3.5Size=78B2026.0386.5
GLM-4.5VSize=106B(A12B)2026.0384.1
GPT 5.2Size=Closed2026.0384
InternVL3.5Size=38B2026.0383.8
Qwen3-VLSize=8B2026.0383.1
Qwen2.5-VLSize=7B2026.0382.6
MolmoSize=72B2026.0381.9
GLM-4.1VSize=9B2026.0380.3
Qwen3-VLSize=4B2026.0380.3
Qwen2.5-VL-7BModel Category=Specialist VLMs2026.0480.3
InternVL3.5Size=8B2026.0379.1
Q-MaskSize=3B2026.0378.7
InternVL3.5Size=20B2026.0378.1
InternVL3.5Size=4B2026.0378
Qwen2.5-VLSize=3B2026.0377.1
TokenVL-8BSize=8B2026.0375.3
MarternSize=8B2026.0375.2
Q-MaskSize=2B2026.0374.4
Claude-3.5-SonnetSize=Closed2026.0374.3
Qwen3-VLSize=2B2026.0372.4
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+23.1, Input Res.=2048, #Visual Tokens=1280, Vis. Enc. Size(M)=125, TTFT (ms)=37212024.1271.1
LLaDA2.0-UniModel Category=Unified Models2026.0470.1
VideoLLaMA3Architecture Category=Token Insertion – Proprietary2025.1269.4
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+12.5, Input Res.=2048, #Visual Tokens=1280, Vis. Enc. Size(M)=125, TTFT (ms)=37212024.1268.3
InternVL-UModel Category=Unified Models2026.0468.3
TextHawk2Size=7B2026.0367.8
GPT-4oSize=Closed2026.0366.4
LLaDA-VModel Category=Specialist VLMs2026.0466.3
Qwen2-VL# train tokens=1.4T, Architecture Category=Token Insertion – Proprietary2025.1265.5
Gemma3Size=27B2026.0365.1
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+12.5, Input Res.=2048, #Visual Tokens=1280, Vis. Enc. Size(M)=125, TTFT (ms)=12632024.1262.5
InsertionHe-2B# train tokens=0.1T, Architecture Category=Token Insertion – Public data, LLM Size=2B2025.1261.8
HunyuanOCRSize=0.9B2026.0361.6
PDF-WuKong2024.1061.3
TokenVL-2BSize=2B2026.0361
InternVL2.5# train tokens=0.5T, Architecture Category=Token Insertion – Proprietary2025.1260.9
IXC2-4KHD-162024.1060.8
BAGELModel Category=Unified Models2026.0460.7
DAVELLM Backbone=Qwen-2.5-7B-Instruct2025.1260.2
Mini-MonkeySize=2B2026.0360.1
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+12.5, Input Res.=2048, #Visual Tokens=1280, Vis. Enc. Size(M)=125, TTFT (ms)=12632024.1260
AIMv2LLM Backbone=Qwen-2.5-7B-Instruct2025.1257.8
Gemini 3.0 ProSize=Closed2026.0357.2
Claude-3-OpusSize=Closed2026.0355.6
SigLIP 2LLM Backbone=Qwen-2.5-7B-Instruct2025.1255.2
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+23.1, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=6412024.1255.1
LLaDA-oModel Category=Unified Models2026.0454.7
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+23.1, Input Res.=2048, #Visual Tokens=1280, Vis. Enc. Size(M)=125, TTFT (ms)=9182024.1253.9
CREAM2024.1053.6
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+12.5, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=6412024.1251.2
DocOwl-1.5-ChatSize=8B2026.0350.7
TextHawk2024.1050.6
DAVELLM Backbone=Llama-3.2-3B-Instruct2025.1250.2
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+11.9, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=6412024.1249.7
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+23.1, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=2332024.1249.6
PixtralSize=12B2026.0349.5
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+12.5, Input Res.=2048, #Visual Tokens=1280, Vis. Enc. Size(M)=125, TTFT (ms)=9182024.1249.3
CASA→# train tokens=0.3T, Architecture=CASAHe-2B, CASA design variant=→, LLM Size=2B2025.1248.6
CASA⊕# train tokens=0.3T, Architecture=CASAHe-2B, CASA design variant=⊕, LLM Size=2B2025.1248.2
DocKylinSize=7B2026.0346.6
DocOwl 22024.1046.4
LLAVAOVVision Encoder=ViT-SO400M, LLM=Qwen2, Data (M) (PT+IT)=4.5+3.2, Input Res.=1152, #Visual Tokens=7290, Vis. Enc. Size(M)=430, TTFT (ms)=141242024.1246.3
ILLUMELLM=Vicuna-7B, Category=Unify Understanding and Generation2024.1245.5
MM1Vision Encoder=ViT-H, LLM=3000+1.5, Input Res.=1344, #Visual Tokens=720, Vis. Enc. Size(M)=632, TTFT (ms)=-2024.1245.5
CASA∨# train tokens=0.3T, Architecture=CASAHe-2B, CASA design variant=∨, LLM Size=2B2025.1244.8
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+11.9, Input Res.=768, #Visual Tokens=144, Vis. Enc. Size(M)=125, TTFT (ms)=4462024.1244.3
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+6.5, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=6412024.1244.2
Emu3-ChatLLM=8B from scratch, Category=Understanding Only2024.1243.8
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+23.1, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=1662024.1243.3
SmolVLMArchitecture Category=Token Insertion – Public data, LLM Size=2B2025.1242.2
UReaderSize=7B2026.0342.2
DophineLLM Backbone=Qwen-2.5-7B-Instruct2025.1242
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+12.5, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=2332024.1241.7
SigLIP 2LLM Backbone=Llama-3.2-3B-Instruct2025.1240.6
MM1Vision Encoder=ViT-H, LLM=3000+1.5, Input Res.=1344, #Visual Tokens=720, Vis. Enc. Size(M)=632, TTFT (ms)=-2024.1238.5
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+1.1, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=6412024.1237.5
DophineLLM Backbone=Llama-3.2-3B-Instruct2025.1237.4
LLaVA-NeXTLLM=Vicuna-7B, Category=Understanding Only2024.1237.1
mPLUG-Owl3 8B# train tokens=0.1T, Architecture Category=Cross-attention-based – Public data, LLM Size=8B2025.1236.8
LLaMA3.2Size=11B2026.0336.6
MonkeySize=10B2026.0336.1
Monkey2024.1036.1
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+12.5, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=1662024.1235.8
AIMv2LLM Backbone=Llama-3.2-3B-Instruct2025.1235.6
Qwen-VL2024.1035.4
FastVLMVision Encoder=FastViTHD, LLM=Vicuna, Data (M) (PT+IT)=15+1.1, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=5772024.1234.7
FastVLMVision Encoder=FastViTHD, LLM=Qwen2, Data (M) (PT+IT)=15+1.1, Input Res.=768, #Visual Tokens=144, Vis. Enc. Size(M)=125, TTFT (ms)=4462024.1234.3
FastVLMVision Encoder=FastViTHD, LLM=Vicuna, Data (M) (PT+IT)=15+1.1, Input Res.=768, #Visual Tokens=144, Vis. Enc. Size(M)=125, TTFT (ms)=3872024.1232
FastVLMVision Encoder=FastViTHD, LLM=Vicuna, Data (M) (PT+IT)=0.5+1.1, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=5772024.1232
FastVLMVision Encoder=FastViTHD, LLM=Vicuna, Data (M) (PT+IT)=0.5+1.1, Input Res.=768, #Visual Tokens=144, Vis. Enc. Size(M)=125, TTFT (ms)=3872024.1229.7
FastVLMVision Encoder=FastViTHD, LLM=Vicuna, Data (M) (PT+IT)=0.5+0.6, Input Res.=1024, #Visual Tokens=256, Vis. Enc. Size(M)=125, TTFT (ms)=5772024.1228.9
FastVLMVision Encoder=FastViTHD, LLM=Vicuna, Data (M) (PT+IT)=0.5+0.6, Input Res.=768, #Visual Tokens=144, Vis. Enc. Size(M)=125, TTFT (ms)=3872024.1228.7
MAE-WebLLM Backbone=Qwen-2.5-7B-Instruct2025.1228.6
TextMonkeySize=8B2026.0328.6
Text-Monkey2024.1028.6
mPLUG-Owl3 2B# train tokens=0.1T, Architecture Category=Cross-attention-based – Public data, LLM Size=2B2025.1228.1
MAE-WebLLM Backbone=Llama-3.2-3B-Instruct2025.1226.2
LLaVA-1.5LLM=Vicuna-7B, Category=Understanding Only2024.1225.8