Unitree CEO Wang Xingxing says a GPT-7 model could directly crack the problems of robotic AI within a year or two, signaling a potential leap for embodied intelligence.
Unitree Chief Executive Wang Xingxing says a future GPT-7 could deliver a major leap in robotics, specifically embodied intelligence. He suggests breakthroughs could arrive within one to two years, building on OpenAI’s GPT-6 Astra progress.
What this could mean for you: robots that understand you better and can follow simple language commands in new environments without heavy reprogramming. A test at Stanford combining a Unitree robot with GPT-6 Astra showed promising results in a kitchen-cleaning scenario, hinting at practical progress rather than abstract theory.
Unitree is laying down a bold forecast: a GPT-7 model could directly crack the stubborn problems of embodied intelligence in robotics within the next year or two. The claim comes from Wang Xingxing, the company’s chief executive, who spoke at an industry event in Shanghai (as reported by media covering the event). He suggests that the upcoming GPT-7 could bridge the gap between AI software and physical robots, enabling more reliable, instruction-following behavior in real-world settings.
Wang’s optimism rests on the momentum around OpenAI’s GPT-6 Astra, which he says has driven the belief that meaningful breakthroughs in robotics are possible sooner rather than later. He cites a benchmark in which robots could complete 80% of assigned tasks in 80% of unfamiliar environments using natural language or voice commands, a threshold he believes could be reachable with GPT-7’s advances.
Unitree’s push isn’t just about smarter chatty AI. The company is focusing on software that enables its robots to understand and act on human intent in the real world, reducing the need for task-specific training data. This aligns with broader industry chatter about an upcoming leap in embodied intelligence, where AI not only reasons in software but also physically interprets and acts in the world.
Despite the enthusiasm, there are caution flags. OpenAI’s models have shown powerful capabilities, but experts warn about the risks if powerful AI systems act unpredictably, especially in robotics where safety is at stake. Wang’s comments brushed past specific technical details and potential safety controls, focusing instead on the potential for a “ChatGPT moment” in robotics within the next couple of years.
In plain terms, embodied intelligence means a robot can understand a command, reason about the environment, and carry out a task without extensive hand-holding or custom-programmed rules for every situation. A GPT-7 that reliably collaborates with hardware could cut the time to program and deploy robots in new settings, from warehouses to homes, by letting people instruct robots with simple language and get predictable results.
The discussion leans on progress shown with GPT-6 Astra, which the leader referencing benchmarks like the ScreenSpot-Pro test, where an AI model’s ability to interact with interfaces is measured. While the exact scores and tests aren’t spelled out in every account, the gist is that improved AI models translate into smarter, more adaptable robotic behavior.
As of now, this is a prediction tied to a future AI model. It highlights a trend you may feel in daily life years from now: smoother voice-powered robot control and fewer specialized instructions to teach a robot a new job. It doesn’t mean consumer robots will suddenly become self-aware or risk-prone overnight. It signals a potential step change in how quickly robots can learn from natural language and adapt to unfamiliar environments.
If GPT-7 advances follow the pace described by Unitree’s CEO, you might see more capable service robots and industrial bots that can be taught with spoken language rather than specialized programs. Exact release timelines, pricing, and device compatibility remain unconfirmed, and OpenAI has not publicly detailed GPT-7’s specifications yet.
The CEO said a GPT-7 model released next year could directly crack problems of embodied intelligence, potentially enabling robots to perform tasks in unfamiliar environments via voice or text commands.
No. It’s a forward-looking claim from Unitree’s leader based on expectations around GPT-7 and GPT-6 Astra benchmarks, not a company-wide product announcement.
You're reading the quick version.
Read next
Rubedo anti-aging skin drug cuts skin age by about two years
A topical cream targeting cellular senescence lowers a skin aging clock in a small study, with potential for broader age-related benefits.
At Europe’s flagship customer contact event, AI dominates discussions and Nebi shows how it’s turning conversations into data-driven actions.
A so-called always-on AI agent ran dozens of autonomous tasks, spending hundreds of thousands of tokens in a single day, raising questions about cost and control.
As AI agents take on more tasks, they raise questions about privacy, consent, and what happens if they act without permission.
The Daily Brief
Every morning, the news that matters from AI, phones, apps and the people shaping tech. Explained in plain English. Free.
One email a day. No spam, unsubscribe anytime. Privacy policy
