RoboVista: Evaluating Vision-Language Models for Diverse Robot ApplicationsRSS 2026
Best overall
Best open-weight · new
Models evaluated
Sorted by standard-prompt accuracy. CoT helps the Gemma 4 family (+3 to +4), the thinking-tuned variants and the Qwen3.6 dense model; it costs accuracy on most other open models.
Standard-prompt runs. Domains follow the benchmark's taxonomy; Driving and Surgery are the smallest columns (n=20, n=32), so read those with wider error bars in mind.
Standard-prompt runs, split by ability type (SC n=306, PL n=148 — planning here bundles decision-making, motion awareness and recovery). Gemini holds its accuracy on planning questions; nearly every other model drops.
Exact figures for every completed pass.
| model | weights | standard | cot | Δ | SC | PL | released |
|---|