Multimodal Deep Search on HR-MMSrch
54.43AccuracySenseNova-MARS-32B
Evaluation results
| Method | Links | |
|---|---|---|
| SenseNova-MARS-32BEvaluation Protocol=Multimodal Deep Search Agents, Model Scale=32B2026.04 | 54.43 | |
| MTA-DeepSearch-32BEvaluation Protocol=Ours, Model Scale=32B2026.04 | 53.95 | |
| Gemini-3 ProEvaluation Protocol=Agent Workflow2026.04 | 53.2 | |
| GPT-5Evaluation Protocol=Agent Workflow2026.04 | 52.13 | |
| Gemini-2.5 ProEvaluation Protocol=Agent Workflow2026.04 | 48.2 | |
| MTA-DeepSearch-8BEvaluation Protocol=Ours, Model Scale=8B2026.04 | 47.54 | |
| SenseNova-MARS-8BEvaluation Protocol=Multimodal Deep Search Agents, Model Scale=8B2026.04 | 41.64 | |
| Qwen3-VL-32B-Inst.Evaluation Protocol=Agent Workflow, Model Scale=32B2026.04 | 38.69 | |
| Qwen3-VL-8B-Inst.Evaluation Protocol=Agent Workflow, Model Scale=8B2026.04 | 32.13 | |
| Gemini-3-proEvaluation Protocol=Direct Answer2026.04 | 23.61 | |
| MMSearch-R1-7BEvaluation Protocol=Multimodal Deep Search Agents, Model Scale=7B2026.04 | 20.33 | |
| GPT-5Evaluation Protocol=Direct Answer2026.04 | 16.39 | |
| Gemini-2.5-proEvaluation Protocol=Direct Answer2026.04 | 15.41 | |
| Qwen3-VL-8B-Inst.Evaluation Protocol=Direct Answer, Model Scale=8B2026.04 | 4.59 | |
| Qwen3-VL-32B-Inst.Evaluation Protocol=Direct Answer, Model Scale=32B2026.04 | 4.59 |