List your company
ResearchFormerly WizWand

Real-world Visual Question Answering on RealWorldQA

79.35AccuracyQwen 3.5
64.11468.069572.02575.9805Dec 5, 2024Mar 9, 2025Jun 12, 2025Sep 14, 2025Dec 18, 2025Mar 22, 2026Jun 25, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Qwen 3.5Zero-shot=true, Parameters=122B-A10B, Organization=Alibaba Cloud2026.0579.35——
Qwen3-VL-4B-MastersSize=4B2025.1277.8——
Qwen3-VL-8B-MastersSize=8B2025.1277.5——
Qwen2.5-VL-7B-MastersSize=7B2025.1277.3——
GPT-4o2024.1275.4——
InternVL3.5-8B-MastersSize=8B2025.1274.9——
InternVL3.5-38B + LoTModel Scale=Large, Framework=LoT, Visual Highlighting=true2026.0474.9——
InternVL3-8B-MastersSize=8B2025.1274.8——
BAGEL (Ours)Baseline=BAGEL, Params=7B active2026.0673.9——
Keye-VL-1.5-8BSize=8B2025.1273.5——
Ovis2-8BSize=8B2025.1272.5——
InternVL3.5-38BModel Scale=Large, Framework=Baseline, Visual Highlighting=false2026.0472.5——
Qwen2.5vl-InstructModel Scale=7B, Training Protocol=PSO, Category=Open-Source Reasoning MLLMs2025.1272.3——
LLaVA-OneVision# Parameters=72B2024.1271.9——
LLaVA-OneVisionModel Scale=72B, Category=Open-Source General MLLMs2025.1271.9——
InternVL2# Parameters=40B2024.1271.8——
CoVT2026.0271.6——
Qwen3-VL-4B + LoTModel Scale=Small, Framework=LoT, Visual Highlighting=true2026.0471.6——
GLM-4.5VZero-shot=true, Parameters=106B-A12B, Organization=Z.ai2026.0571.38——
BAGELBaseline=BAGEL, Params=7B active2026.0671.2——
Ovis2-4BSize=4B2025.1271.1——
VARGPT-v1.1 (Ours)Baseline=VARGPT-v1.1, Params=7B+2B2026.0671.1——
BLIP3o-8B (Ours)Baseline=BLIP3o-8B, Params=8B2026.0670.9——
Llama 4 MaverickZero-shot=true, Parameters=400B-A17B, Organization=MetaAI2026.0570.85——
InternVL3-8BSize=8B2025.1270.8——
Qwen2.5-VL-3B-MastersSize=3B2025.1270.7——
GLM-4.1V-9BSize=9B2025.1270.6——
Qwen3-VL-4BSize=4B2025.1270.6——
CrystaLBase Model=Qwen2.5-VL-7B2026.0270.6——
Baseline (No Recompute)Model=Qwen3-VL-8B, Recomputation budget (k)=02026.0370.59——
SKILA2026.0270.5——
Gemini 1.5 Pro2024.1270.4——
Qwen2.5vl-InstructModel Scale=7B, Training Protocol=SFT + GRPO, Category=Open-Source Reasoning MLLMs2025.1270.2——
Qwen2-VL# Parameters=8B2024.1270.1——
LLaVA-OneVision-7BSize=7B2025.1269.9——
Qwen3-VL-8BSize=8B2025.1269.9——
Qwen-2.5-VL-7B + EvoLMMsetting=zero-shot2025.1169.9——
NVLM-D-1.0# Parameters=78B2024.1269.7——
GPT-4o2026.0269.7——
NVILA# Parameters=15B2024.1269.5——
LIVR2026.0269.2——
MODIXSize=8B, Backbone=Qwen3-VL2026.0469.15——
Qwen3-VL-2B-MastersSize=2B2025.1269——
BLIP3o-8BBaseline=BLIP3o-8B, Params=8B2026.0669——
MODIXSize=3B, Backbone=LFM2-VL2026.0468.76——
Qwen-2.5-VL-7Bsetting=zero-shot2025.1168.7——
NVILA# Parameters=8B2024.1268.6——
Qwen2.5-VL-7BParameters=7B2026.0268.6——
Qwen2.5vl-InstructModel Scale=7B, Category=Open-Source Reasoning MLLMs2025.1268.5——
Qwen2.5-VL-7BSize=7B2025.1268.5——
InternVL3-2B-MastersSize=2B2025.1268.4——
LLaVA-OneVision-1.5-8BSize=8B2025.1268.1——
InternVL3.5-4B-MastersSize=4B2025.1268.1——
InternVL3.5-2B-MastersSize=2B2025.1268.1——
InfoFlow KVModel=Qwen3-VL-8B, Recomputation budget (k)=22026.0368.1——
MiniCPM-o2.6-8BSize=8B2025.1268——
Cambrian-1# Parameters=34B2024.1267.8——
LLaVA-OneVision-1.5-4BSize=4B2025.1267.8——
Keye-VL-8BSize=8B2025.1267.7——
LVR2026.0267.7——
CacheBlendModel=Qwen3-VL-8B, Recomputation budget (k)=22026.0367.58——
InternVL3.5-8BSize=8B2025.1267.5——
Qwen2.5-VL-7B + LoTModel Scale=Medium, Framework=LoT, Visual Highlighting=true2026.0467.5——
Qwen3-VL-8B + LoTModel Scale=Medium, Framework=LoT, Visual Highlighting=true2026.0467.5——
VARGPT-v1.1Baseline=VARGPT-v1.1, Params=7B+2B2026.0667.5——
No RecomputeModel=Qwen3-VL-8B, Recomputation budget (k)=22026.0367.45——
EPICModel=Qwen3-VL-8B, Recomputation budget (k)=22026.0367.45——
LFM2-VLSize=3B, Configuration=Baseline2026.0467.32——
Qwen2.5-VL-7BRatio=100%, Backbone=Qwen2.5-VL-7B2026.0567.32——
InternVL3.5-4B + LoTModel Scale=Small, Framework=LoT, Visual Highlighting=true2026.0467.3——
NVILA-Lite# Parameters=15B2024.1267.1——
Qwen2.5-VL-32B + LoTModel Scale=Large, Framework=LoT, Visual Highlighting=true2026.0467.1——
Qwen3-VLSize=8B, Configuration=Baseline2026.0466.93——
Qwen3-VL-8BModel Scale=Medium, Framework=Baseline, Visual Highlighting=false2026.0466.8——
Qwen3-VL-4BModel Scale=Small, Framework=Baseline, Visual Highlighting=false2026.0466.7——
InternVL3.5-4BModel Scale=Small, Framework=Baseline, Visual Highlighting=false2026.0466.7——
InfoFlow KVModel=Qwen3-VL-8B, Recomputation budget (k)=42026.0366.67——
V-GIFTModel=LLaVA-OneVision-1.52026.0466.4——
LLaVA-OneVision# Parameters=8B2024.1266.3——
LLaVA-OneVisionModel Scale=7B, Category=Open-Source General MLLMs2025.1266.3——
InternVL3.5-4BSize=4B2025.1266.3——
Phoenix-VL 1.5 MediumZero-shot=true, Parameters=123B, Organization=HTX2026.0566.27——
BaselineModel=LLaVA-OneVision-1.52026.0466.1——
Ovis2-2BSize=2B2025.1266——
InternVL3.5-8B + LoTModel Scale=Medium, Framework=LoT, Visual Highlighting=true2026.0465.9——
No RecomputeModel=Qwen3-VL-8B, Recomputation budget (k)=42026.0365.88——
MODIXSize=2B, Backbone=Qwen3-VL2026.0465.75——
CacheBlendModel=Qwen3-VL-8B, Recomputation budget (k)=42026.0365.62——
NVILA-Lite# Parameters=8B2024.1265.6——
Qwen2.5-VL-32BModel Scale=Large, Framework=Baseline, Visual Highlighting=false2026.0465.6——
EPICModel=Qwen3-VL-8B, Recomputation budget (k)=42026.0365.49——
Pixtral# Parameters=12B2024.1265.4——
Qwen2.5-VL-3BSize=3B2025.1265.4——
Qwen2VL-7B-Instruct + LUVCLatency=0.2335, Speedup=1.53x2025.1265.2376.52—
Qwen2VL-7B-InstructLatency=0.35692025.1265.177.38—
MiniCPM-V2.6-8BSize=8B2025.1265——
Aquila-VL-2BSize=2B2025.1265——
Qwen2.5-VL-7BModel Scale=Medium, Framework=Baseline, Visual Highlighting=false2026.0465——
Qwen2-VL-7B + LoTModel Scale=Medium, Framework=LoT, Visual Highlighting=true2026.0464.8——
InternVL3.5-8BModel Scale=Medium, Framework=Baseline, Visual Highlighting=false2026.0464.7——