Claude Fable 5.2 scores 64.9% on RoboDojo, more than double GPT-6 Astra on Anthropic's own run
Anthropic's new general model steered a simulated two-arm robot to a 64.9% average success rate, ahead of every frontier rival it tested. On an independent real-arm test in October, the order among those rivals was very different.
2 min readSources verified