Visual Understanding on V* Bench, HR-Bench, and MME RealWorld
85.9Average ScoreSenseNova-MARS-32B
Evaluation results
| Method | Links | |
|---|---|---|
| SenseNova-MARS-32BMethod Category=Agentic Model2025.12 | 85.9 | |
| Skywork-R1V4Method Category=Agentic Model2025.12 | 80.5 | |
| SenseNova-MARS-8BMethod Category=Agentic Model2025.12 | 80.4 | |
| Mini o3Method Category=Agentic Model2025.12 | 76.1 | |
| Qwen3-VL-8B-InstructMethod Category=Direct Answer2025.12 | 75.5 | |
| DeepEyesV2Method Category=Agentic Model2025.12 | 74.6 | |
| ThymeMethod Category=Agentic Model2025.12 | 74 | |
| DeepEyesMethod Category=Agentic Model2025.12 | 72.5 | |
| Pixel-ReasonerMethod Category=Agentic Model2025.12 | 71.9 | |
| Qwen2.5-VL-32B-InstructMethod Category=Direct Answer2025.12 | 68.2 | |
| Qwen2.5-VL-7B-InstructMethod Category=Direct Answer2025.12 | 64.9 | |
| LLaVA-onevisonMethod Category=Direct Answer2025.12 | 63.9 | |
| GPT-4oMethod Category=Direct Answer2025.12 | 63.7 |