Unitree CEO Says a GPT-7 Release Next Year Could 'Directly Crack' Robotic AI
Unitree CEO Wang Xingxing said at a Shanghai industry event that a GPT-7 model released next year could solve embodied intelligence, shortening a timeline he had previously put at a decade.
Speaking at an industry event in Shanghai, Wang said OpenAI's recently released GPT-6 Astra model was powerful enough to support the belief that a breakthrough in robotic AI could arrive "within one or two years," TechRadar reported, citing the South China Morning Post. "It's possible that, say, a GPT-7 model released next year could directly crack the problems of embodied intelligence," Wang said, referring to the use of artificial intelligence to command a robotic device.
The comment marked a change from August, when Wang said such a "ChatGPT moment" for robotics could take a decade to arrive.
Wang also set out a benchmark for judging when embodied intelligence has matured: robots should be able to complete 80% of the tasks assigned to them in 80% of unfamiliar environments using voice or text commands alone. If robots reach that level, he said, embodied intelligence will have crossed its threshold level of maturity.
Unitree's robots are among the most advanced in the world, but many of their strongest abilities lie in their hardware. The company has begun to put more emphasis on AI software, and the progress made in GPT-6 Astra could close the gap between the two sides of its machines.
Part of that combination has already been tested. Stanford University's Movement Lab paired a Unitree robot with GPT-6 Astra and tasked it with cleaning a kitchen it had never seen before. According to the experiment, the robot was able to "clean up across the room and retrieve a remembered object from an underspecified request, without environment-specific training data or additional policy learning."
GPT-6 Astra included significant improvements over earlier models, scoring 92.7% on the ScreenSpot-Pro test, which measures an AI model's ability to interact with interfaces, up from 76.9% for GPT-5.6 Sol. The model has also drawn concern from cybersecurity experts about the threat it could pose if it goes rogue and starts hacking other organizations, behavior that OpenAI's recent models have been known to display.
If Wang's forecast holds and AI capabilities expand sharply over the next year or two, the consequences would reach the robotics industry and other areas of life that AI already touches. Large sums have been invested in AI over recent years, leaving a great deal riding on the bet.