High-resolution perception on VStar
88AccuracyMAESTRO
Evaluation Results
| Method | Links | |
|---|---|---|
| MAESTROModel Type=Ours2026.05 | 88 | |
| DeepEyesModel Type=Think with Images Methods2026.05 | 85.6 | |
| ThymeModel Type=Think with Images Methods2026.05 | 82.2 | |
| Chain-of-FocusModel Type=Think with Images Methods2026.05 | 82.2 | |
| DeepEyes-v2Model Type=Think with Images Methods2026.05 | 81.8 | |
| PixelReasonerModel Type=Think with Images Methods2026.05 | 81.7 | |
| GLM-4.6VModel Type=Open-Source & Baselines2026.05 | 81.2 | |
| Gemini-2.5-ProModel Type=Closed-Source2026.05 | 79.1 | |
| VTOOL-R1Model Type=Think with Images Methods2026.05 | 78.5 | |
| Qwen3-VL-32BModel Type=Open-Source & Baselines2026.05 | 78 | |
| MathCoder-VLModel Type=Think with Images Methods2026.05 | 77.5 | |
| Direct AnsweringModel Type=Open-Source & Baselines2026.05 | 77 | |
| VTS-VModel Type=Think with Images Methods2026.05 | 75.9 | |
| Kimi-K2.5Model Type=Open-Source & Baselines2026.05 | 72.8 | |
| GPT-5Model Type=Closed-Source2026.05 | 72.5 | |
| Gemini-2.5-FlashModel Type=Closed-Source2026.05 | 72.3 | |
| GPT-4oModel Type=Closed-Source2026.05 | 66 | |
| VisionReasonerModel Type=Think with Images Methods2026.05 | 59.7 | |
| Visual-ARFTModel Type=Think with Images Methods2026.05 | 58.6 | |
| Untrained ModelModel Type=Open-Source & Baselines2026.05 | 41.4 |