All Frontier AI Models Fail Humanoid Tasks in First Robotics Benchmark

Artificial Analysis launched AA-Robotics v0.1, a research preview benchmarking language models on physical tasks using a 6-axis robot arm. GPT-6 Astra led with a score of 69, followed by GPT-6.1 Sol at 65 and Gemini 4 Argon at 55, though all models were roughly 100x slower and 10x more expensive than human baselines. No model scored above 2/100 on 11 humanoid tasks requiring balance and walking, with GPT-6 models initiating movement after 75 to 80 seconds and Grok 4.7 and Gemini 4 Argon failing to move the robot in most episodes.