List your company
ResearchFormerly WizWand

Multimodal Search-based Question Answering on MMSearch

69.6AccuracyVision-DeepResearch-8B
10.52825.86441.256.536Aug 31, 2025Oct 15, 2025Nov 30, 2025Jan 15, 2026Mar 1, 2026Apr 16, 2026Jun 1, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Vision-DeepResearch-8BWorkflow=Agent Workflow2026.0169.6
Vision-DeepResearch-30B-A3BWorkflow=Agent Workflow2026.0169.6
Gemini-2.5 ProWorkflow=Agent Workflow2026.0169
Claude-4-SonnetWorkflow=Agent Workflow2026.0167.2
GoG-3-8B-Think-RLEvaluation Methodology=Search-Equipped Models2026.0165.5
Gemini-2.5 FlashWorkflow=Agent Workflow2026.0164
Qwen3-VL-8B-ThinkingEvaluation Methodology=Full-Search Workflow2026.0163.84
GPT-5Workflow=Agent Workflow2026.0163.7
Claude-3.7-SonnetWorkflow=Agent Workflow2026.0163.7
DeepEyesV2Tool=General, Model Size=7B2025.1163.7
DeepEyesV2Backbone=Qwen2.5-VL-7B2025.0863.7
GoG-2.5-7B-RLEvaluation Methodology=Search-Equipped Models2026.0156.73
Webwatcher-32BWorkflow=Multimodal DeepResearch MLLM2026.0155.3
Qwen3-VL-30B-A3B-InstructWorkflow=Agent Workflow2026.0155
PMSRBackbone=Qwen2.5-VL-7B2025.0854.3
MMSearch-R1-7BWorkflow=Multimodal DeepResearch MLLM2026.0153.8
GoG-3-8B-Think-SFTEvaluation Methodology=Search-Equipped Models2026.0153.8
MMSearch-R1Tool=Search, Model Size=7B2025.1153.8
MMSearch-R1Backbone=Qwen2.5-VL-7B2025.0853.8
Qwen3-VL-30B-A3B-ThinkingWorkflow=Agent Workflow2026.0153.2
Qwen-2.5-VLTool=Search, Model Size=7B2025.1152.2
Qwen3-VL-8B-InstructWorkflow=Agent Workflow2026.0152
OmniSearchBackbone=GPT-4o2025.0849.7
Webwatcher-7BWorkflow=Multimodal DeepResearch MLLM2026.0149.1
WebWatcherTool=Search, Model Size=7B2025.1149.1
WebWatcherBackbone=Qwen2.5-VL-7B2025.0849.1
Qwen3-VL-8B-ThinkingWorkflow=Agent Workflow2026.0145.6
Gemini-2.5-flashWorkflow=RAG Workflow2026.0143.9
MMSearch-R1*Evaluation Methodology=Search-Equipped Models2026.0140.94
GoG-2.5-7B-SFTEvaluation Methodology=Search-Equipped Models2026.0140.35
Gemini-2.5 ProWorkflow=Direct Answer2026.0139.8
Qwen3-VL-8B-ThinkEvaluation Methodology=Prompt-based GoG Method2026.0136.84
Qwen2.5-VL-7BEvaluation Methodology=Full-Search Workflow2026.0136.26
GPT-5Workflow=Direct Answer2026.0133.3
Claude-3.7-SonnetWorkflow=RAG Workflow2026.0132.7
Gemini-2.5 FlashWorkflow=Direct Answer2026.0130.4
Qwen-2.5-VL-72BWorkflow=RAG Workflow2026.0129.2
Gemini 2.5 ProTool=✗, Model Size=-2025.1126.9
GPT4oTool=✗, Model Size=-2025.1122.2
GPT-4oEvaluation Methodology=Direct Answer2026.0121.64
PIXELRAGReader=Qwen3-VL-4B, Top-k=3, Retrieval Category=Pixel-based retrieval2026.0621.6
Claude-3.7-SonnetWorkflow=Direct Answer2026.0121.1
Qwen2.5-VL-7BEvaluation Methodology=Prompt-based GoG Method2026.0120.47
PIXELRAG (base)Reader=Qwen3-VL-4B, Top-k=3, Retrieval Category=Pixel-based retrieval2026.0619.9
Qwen3-VL-30B-A3B-ThinkingWorkflow=Direct Answer2026.0119.3
Claude-4-SonnetWorkflow=Direct Answer2026.0118.7
Qwen3-VL-30B-A3B-InstructWorkflow=Direct Answer2026.0118.7
Qwen3-VL-30B-A3B-InstructEvaluation Methodology=Direct Answer2026.0118.13
Qwen3-VL-30B-A3B-ThinkingEvaluation Methodology=Direct Answer2026.0117.54
Qwen2.5-VL-32BEvaluation Methodology=Direct Answer2026.0116.37
Qwen3-VL-8B-ThinkingWorkflow=Direct Answer2026.0115.8
Qwen3-VL-8B-InstructWorkflow=Direct Answer2026.0115.2
Qwen3-VL-8B-InstructEvaluation Methodology=Direct Answer2026.0115.2
Qwen3-VL-8B-ThinkingEvaluation Methodology=Direct Answer2026.0115.2
Qwen2.5-VL-7BEvaluation Methodology=Direct Answer2026.0114.62
No retrievalReader=Qwen3-VL-4B, Top-k=3, Retrieval Category=None2026.0613.5
TrafilaturaReader=Qwen3-VL-4B, Top-k=3, Retrieval Category=Text-based retrieval2026.0613.5
mwparserfromhellReader=Qwen3-VL-4B, Top-k=3, Retrieval Category=Text-based retrieval2026.0612.9
Qwen-2.5-VLTool=✗, Model Size=7B2025.1112.8