Embodied Visual Agent Task on VisualAgentBench (test)
58.65AVG. ScoreReCAPA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ReCAPAModel Category=Proprietary-LMMs, Evaluation Protocol=Prompting2026.04 | 58.65 | 50.6 | 66.7 | |
| GPT-4o miniModel Category=Proprietary-LMMs, Evaluation Protocol=Prompting2026.04 | 54.15 | 46.7 | 61.6 | |
| Gemini 2.5 FlashModel Category=Proprietary-LMMs, Evaluation Protocol=Prompting2026.04 | 53 | 43.9 | 62.1 | |
| Claude-4-SonnetModel Category=Proprietary-LMMs, Evaluation Protocol=Prompting2026.04 | 50.25 | 42.6 | 57.9 | |
| GPT-4oModel Category=Proprietary-LMMs, Evaluation Protocol=Prompting2026.04 | 48.3 | 41.4 | 55.2 | |
| GPT-4V (preview)Model Category=Proprietary-LMMs, Evaluation Protocol=Prompting2026.04 | 41.95 | 36.5 | 47.4 | |
| Claude-3.5-SonnetModel Category=Proprietary-LMMs, Evaluation Protocol=Prompting2026.04 | 40.15 | 24.3 | 56 | |
| InternVL-2Model Category=Open-LMMs, Evaluation Protocol=Fine-tuning2026.04 | 22.2 | 16 | 28.4 | |
| LLaVA-NeXTModel Category=Open-LMMs, Evaluation Protocol=Fine-tuning2026.04 | 16.6 | 9.4 | 23.8 | |
| GLM-4VModel Category=Open-LMMs, Evaluation Protocol=Fine-tuning2026.04 | 14.35 | 8.8 | 19.9 | |
| CogVLM2Model Category=Open-LMMs, Evaluation Protocol=Fine-tuning2026.04 | 13.55 | 6.6 | 20.5 | |
| Qwen-VLModel Category=Open-LMMs, Evaluation Protocol=Fine-tuning2026.04 | 9.9 | 1.7 | 18.1 | |
| qwen-vl-maxModel Category=Proprietary-LMMs, Evaluation Protocol=Prompting2026.04 | 2.65 | 0 | 5.3 |