Nvidia CEO Jensen Huang declares ‘AGI has arrived’ after OpenAI launches GPT-6 Astra

Nvidia CEO Jensen Huang has declared that artificial general intelligence, or AGI, has arrived following the launch of OpenAI’s GPT-6 Astra. Huang made the statement on X while congratulating OpenAI on its latest AI development. He highlighted the extensive Nvidia infrastructure used to train Astra, saying the model was trained on more than 100,000 Nvidia Grace Blackwell NVLink 72 systems. Huang also said another 400,000 Nvidia GPUs are expected to come online as AI infrastructure continues to expand. OpenAI introduced GPT-6 Astra last week as its most advanced artificial intelligence model to date. OpenAI President Greg Brockman described the system as a “generational leap” in AI capabilities. AGI generally refers to AI systems capable of performing a broad range of intellectual tasks rather than being restricted to specialized functions. OpenAI’s charter defines AGI as autonomous systems capable of outperforming humans in most economically valuable work. Astra has demonstrated stronger performance on computer-use and task-completion benchmarks. In OpenAI’s OSWorld 2.0 evaluation, Astra reportedly achieved a 72.6% success rate while completing tasks in about 40 minutes on average. By comparison, GPT-5.6 Sol reportedly required around 75 minutes for similar tasks and achieved a slightly higher success rate of 75%. The results highlight Astra’s focus on completing complex tasks more efficiently. Users can provide Astra with a broader objective and allow the system to carry out multiple steps with limited human intervention. The model can browse the internet, operate computers, interact with software, write code and conduct research. It is also designed to make decisions in situations where instructions may be incomplete or ambiguous. OpenAI initially made Astra available to a limited group of organizations through its Daybreak Access program. The model is also being offered to users across ChatGPT Plus, Pro, Business and Enterprise plans, along with access through the OpenAI API and AWS. Despite Huang’s declaration, OpenAI has not officially classified GPT-6 Astra as AGI. Brockman has described the model as a possible early step toward achieving AGI, while OpenAI CEO Sam Altman has previously questioned the usefulness of the term. Altman has characterized AGI as an increasingly irrelevant marketing label rather than a precise technical milestone. Huang’s comments also come as Nvidia and OpenAI deepen their relationship around AI infrastructure. The companies previously announced plans for Nvidia to help develop at least 10 gigawatts of AI data-center capacity for OpenAI, alongside a proposed Nvidia investment of up to $100 billion. Huang later clarified that the $100 billion figure was not a firm financial commitment, and the broader plan subsequently stalled. In February, OpenAI announced a new $30 billion investment from Nvidia as part of a larger funding round. OpenAI also said the companies were expanding their long-standing infrastructure partnership, including dedicated capacity for both AI inference and training on Nvidia’s Vera Rubin systems. The company said the expansion builds on Hopper and Blackwell systems already operating across major infrastructure providers. The developments underscore the growing importance of advanced computing infrastructure as AI companies compete to develop increasingly capable models. Huang’s declaration that AGI has arrived is therefore likely to fuel further debate over whether current frontier AI systems have crossed the threshold traditionally associated with artificial general intelligence.