RoboDojo evaluates GPT-6 Astra as an embodied agent
Compared with GPT-5.5 and DeepSeek-Flash across sim and real robots, humanoid high-level control, dexterous piano playing and in-context learning, Astra shows strong semantic and spatial understanding but weak physical commonsense.
We put GPT-6 Astra in the RoboDojo. 🥋🤖 The RoboDojo Team conducted a comprehensive evaluation of GPT-6 Astra as an embodied agent, including: • RoboDojo Sim & Real, compared with GPT-5.5 and DeepSeek-Flash • Humanoid high-level control • Dexterous piano playing with RoboPianist 🎹 • A systematic study of in-context learning (ICL) Our key takeaway: GPT-6 Astra demonstrates remarkably strong semantic and spatial understanding, together with impressive in-context adaptation. At the same time, physical commonsense remains a clear bottleneck — revealing an important gap between understanding the world and truly reasoning about its physics. Full report & demos: https://robodojo-benchmark.com/report/gpt-6-astra-eval @_wenbozhang (project lead), @wenhaocha1, @frankzydou, @JinWeiyang18434, @YutaoOuyang, @minifullcapsule, @x_h_ucb, @YutaoOuyang, @YueChen614