Visual Reasoning on VSP
83.7AccuracyDeepLatent-RL-7B
Evaluation results
| Method | Links | |
|---|---|---|
| DeepLatent-RL-7BModel=DeepLatent-RL-7B2026.05 | 83.7 | |
| AdaReasoner 7BModel Category=Tool-Planning Models, Backbone=7B2026.01 | 78.36 | |
| Mirage-7BModel=Mirage-7B2026.05 | 76 | |
| GPT 5 + ToolsModel Category=Tool-Planning Models, Tool Usage=true2026.01 | 71.36 | |
| Claude 4 sonnetModel Category=Closed-Source Models2026.01 | 56.27 | |
| GPT 5Model Category=Closed-Source Models2026.01 | 55.64 | |
| Gemini 2.5 flashModel Category=Closed-Source Models2026.01 | 53.55 | |
| Qwen 2.5 VL 72B + ToolsModel Category=Tool-Planning Models, Tool Usage=true, Backbone=72B2026.01 | 45 | |
| Qwen 2.5 VL 72BModel Category=Open-Source Models, Backbone=72B2026.01 | 39.09 | |
| InternVL3 78BModel Category=Open-Source Models, Backbone=78B2026.01 | 35.09 | |
| Qwen 2.5 VL 32BModel Category=Open-Source Models, Backbone=32B2026.01 | 33.91 | |
| Qwen 2.5 VL 7B + ToolsModel Category=Tool-Planning Models, Tool Usage=true, Backbone=7B2026.01 | 30.45 | |
| Qwen 2.5 VL 7BModel Category=Open-Source Models, Backbone=7B2026.01 | 28.09 | |
| Qwen 2.5 VL 3BModel Category=Open-Source Models, Backbone=3B2026.01 | 26.73 | |
| PixelReasonerModel Category=Tool-Planning Models, Unified Evaluation Framework=true2026.01 | 24.55 | |
| Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.05 | 13.5 | |
| DeepEyesModel Category=Tool-Planning Models, Unified Evaluation Framework=true2026.01 | 12.18 |