San Francisco, September 14 — Nvidia chief executive Jensen Huang says artificial general intelligence is here. In a post on X on Tuesday, 8 September, Huang wrote: “AGI has arrived. Congratulations OpenAI team.” He was responding to the launch of OpenAI’s newest ChatGPT model, GPT-6, which the company calls ChatGPT Astra.
The model arrived the previous week, according to Yahoo Finance reporting. OpenAI describes ChatGPT Astra as “state-of-the-art on computer use, browsing, software engineering, cyber security, science and professional work,” and says it can autonomously complete multiple tasks at once.
In a demonstration, the model handled user requests while simultaneously taking on complex work such as creating a video game or preparing a legal document.
A company divided on the words
OpenAI’s own leadership does not speak with one voice on the label.
During a press call to unveil the latest version of ChatGPT, OpenAI president Greg Brockman said, “welcome to the AGI era,” calling the new model a “generational leap in capability.” He added: “If we fast forward a couple years, and we look back and say when was it really that AGI was created, I think it’s going to be about this time, and I think it might be about this model.” Days earlier, OpenAI chief executive Sam Altman struck a different note.
Speaking on the Sources podcast, Altman said AGI was “at best… a very poorly defined term.” He went further: “I was going to say it’s like an irrelevant marketing term.” That split matters, because the definition itself is contested across the industry. OpenAI’s charter frames AGI as “highly autonomous systems that outperforms humans at most economically valuable work.” No universal, agreed-upon standard exists, even as leading AI developers and hardware companies stake out competing positions on whether the technology has crossed the threshold.
Benchmarks rise, skepticism holds
In benchmark tests designed to measure how close an AI model is to AGI, ChatGPT Astra set new high scores. Greg Kamradt, from the ARC Prize Foundation, which develops the tests, described the results as a “meaningful step change in frontier-model performance.” Not everyone is convinced. In response to Huang’s comments, tech consultant Marko Tasic disputed the claims, describing them as a marketing stunt ahead of a potential initial public offering.
“No, it’s not AGI,” Tasic said. “This is just PR before IPO.
Pump and dump, nothing more.” For now, the argument is less about what the model can do than about what to call it. Huang, whose company builds the hardware much of this software runs on, has picked his side. Altman, whose company built the model, has not.
What to watch next is whether OpenAI’s own charter language — autonomous systems that outperform humans at most economically valuable work — gets tested against Astra in public, and whether other lab leaders follow Huang in saying the line has been crossed.
