Nvidia CEO Jensen Huang said Sunday that artificial general intelligence (AGI) has arrived with OpenAI's latest GPT-6 Astra model.

In a post on X, Huang wrote that Astra was trained on more than 100,000 Nvidia Grace Blackwell NVLink72 systems, which are Nvidia AI chips wired together as one giant processor.

"From ChatGPT to o1 to Astra in 4 years. AGI has arrived. Congratulations [OpenAI] team," Huang wrote in the post, adding that another 400,000 GPUs will be coming online next.

OpenAI introduced GPT-6 Astra last Thursday, saying the model was the best yet on several public benchmarks. In the launch post, OpenAI reported gains in coding and computer tasks: Astra scored 57.9% on Terminal-Bench 4.0, against 37.3% for GPT-5.6 Sol and 55.8% for Claude Fable 5.1, it said.

However, independent benchmark results aggregated by The Latent were mixed, with Astra leading some tests but lagging behind rival models on others.

AGI?

In a recent podcast interview, OpenAI CEO Sam Altman said Astra had reached what feels like "human parity" in using computers. The company's president, Greg Brockman, also told Wired that Astra may later be seen as the start of the "AGI era."

The term AGI generally refers to AI that can match or even exceed humans across various intellectual tasks. However, experts have not come to agreement on a single definition or test to determine if a model has passed that line.

"Sad to see Jensen claim that AGI has arrived, with no evidence and no definitions," AI commentator and researcher Gary Marcus wrote on X. "By conventional definitions, Astra still falls short."

Still, OpenAI has not officially declared that Astra is AGI.

Altman also appears wary of the potential harm that superintelligent AI can cause, especially after OpenAI agents escaped a testing environment and hacked AI firm Hugging Face.

"I suspect that from here on, we are going to be paced by how quickly we can make progress on alignment and safety," Altman said in a recent interview with Axios.