List your company
ResearchFormerly WizWand

Multimodal Deep Search on BC-VL

57.6AccuracyGPT-5
18.54828.686538.82548.9635Apr 7, 2026Apr 12, 2026Apr 18, 2026Apr 24, 2026Apr 29, 2026May 5, 2026May 11, 2026
Updated 6mo ago

Evaluation results

MethodLinks
GPT-5Evaluation Setting=Agent Workflow2026.0557.6
MTA-DeepSearch-32BEvaluation Protocol=Ours, Model Scale=32B2026.0453.77
Gemini-3 ProEvaluation Protocol=Agent Workflow2026.0451.78
GPT-5Evaluation Protocol=Agent Workflow2026.0451.63
Claude-3.7-SonnetEvaluation Setting=Agent Workflow2026.0550.4
Gemini-2.5 ProEvaluation Protocol=Agent Workflow2026.0449.5
Claude-4-SonnetEvaluation Setting=Agent Workflow2026.0548.6
Qwen3-VL-30B + ODE-RLEvaluation Setting=Visual-Native Agent Harness, Training Strategy=ODE-RL2026.0546.1
MTA-DeepSearch-8BEvaluation Protocol=Ours, Model Scale=8B2026.0444.36
Gemini-2.5 ProEvaluation Setting=Direct Reasoning2026.0543.1
Qwen3-VL-30B + ODE-SFTEvaluation Setting=Visual-Native Agent Harness, Training Strategy=ODE-SFT2026.0543.1
MM-DeepResearch 32BEvaluation Protocol=Multimodal Deep Search Agents, Model Scale=32B2026.0443
GPT-5Evaluation Setting=Direct Reasoning2026.0542.9
Gemini-2.5 ProEvaluation Setting=Agent Workflow2026.0542.3
Qwen3-VL-8B + ODE-RLEvaluation Setting=Visual-Native Agent Harness, Training Strategy=ODE-RL2026.0541.9
GPT-5Evaluation Protocol=Direct Answer2026.0441.6
Gemini-3-proEvaluation Protocol=Direct Answer2026.0441.35
Gemini-2.5-proEvaluation Protocol=Direct Answer2026.0439.85
Qwen3-VL-8B + ODE-SFTEvaluation Setting=Visual-Native Agent Harness, Training Strategy=ODE-SFT2026.0539.6
Qwen3-VL-32B-Inst.Evaluation Protocol=Agent Workflow, Model Scale=32B2026.0438.69
Gemini-2.5 FlashEvaluation Setting=Agent Workflow2026.0538.1
Gemini-2.5 FlashEvaluation Setting=Direct Reasoning2026.0537.1
Qwen3-VL-8B-Inst.Evaluation Protocol=Agent Workflow, Model Scale=8B2026.0435.89
Qwen3-VL-30BEvaluation Setting=Visual-Native Agent Harness2026.0532.8
Claude-3.7-SonnetEvaluation Setting=Direct Reasoning2026.0532.3
Qwen3-VL-30BEvaluation Setting=Direct Reasoning2026.0529.6
Claude-4-SonnetEvaluation Setting=Direct Reasoning2026.0529.3
Qwen3-VL-30BEvaluation Setting=Agent Workflow2026.0529.3
Webwatcher-32BEvaluation Protocol=Multimodal Deep Search Agents, Model Scale=32B2026.0426.7
WebWatcher-32BEvaluation Setting=Deep Search Agent2026.0526.7
Qwen3-VL-8BEvaluation Setting=Visual-Native Agent Harness2026.0526.1
Qwen3-VL-8BEvaluation Setting=Direct Reasoning2026.0525.1
Qwen3-VL-32B-Inst.Evaluation Protocol=Direct Answer, Model Scale=32B2026.0424.81
Qwen3-VL-8BEvaluation Setting=Agent Workflow2026.0524.3
Webwatcher-7BEvaluation Protocol=Multimodal Deep Search Agents, Model Scale=7B2026.0420.3
WebWatcher-7BEvaluation Setting=Deep Search Agent2026.0520.3
Qwen3-VL-8B-Inst.Evaluation Protocol=Direct Answer, Model Scale=8B2026.0420.05