Article

Cross-model embodied capability evaluation of four frontier large models on RoboColiseum

The paper conducts a cross-model evaluation of GPT-6 Astra, GPT-6.1 Sol, Claude Opus 5.5, and Gemini 4 Argon on RoboColiseum, covering four tracks: instruction following, spatial reasoning, general manipulation, and robustness.

Part of the event: RoboColiseum Releases Four-Model Embodied Evaluation: Opus Tops

This page provides a summary and source link only.