OpenAI reassigns robotics researchers to other teams as general-purpose models might make specialized robotics AI unnecessary

Why it matters

OpenAI is reportedly reassigning its robotics researchers rather than abandoning the field, folding them into Mark Chen's main research group while applied robotics and hardware staff shift to Greg Brockman's Applied organization. The reported rationale is a strategic bet that general-purpose models such as GPT-6 Astra can directly control robots, making specialized robotics models less necessary. Evidence cited includes an Nvidia study in which an Astra-powered agent outperformed the best specialized models on simulated tasks using a simulated DROID robot, and a new benchmark, StationeryBench, where Astra fully completed some desk-object tasks on dual-arm YAM robotic arms while Ai2's MolmoAct2 completed none. Astra also reportedly navigated and cleaned a kitchen on a humanoid robot. The move contrasts with OpenAI's earlier robotics shutdown, which a co-founder attributed to insufficient training data. Context matters here: OpenAI faces financial pressure to focus on its core AI business, where it has reportedly lost ground to Anthropic, even as it reportedly eyes a future public robot demo and everyday robots. Specialized models may still hold advantages in cost, reliability, speed, and energy use.