On Sunday (local time), Jensen Huang made a post on social media, announcing that OpenAI's GPT-6 Astra model had been trained using a cluster composed of roughly 100,000 NVIDIA Grace Blackwell NVLink72 chips. He extended his congratulations to the OpenAI team and proclaimed, "AGI has officially arrived." Huang also noted that the transition from ChatGPT to o1, and subsequently to Astra, had taken a total of four years. Furthermore, he disclosed that the next batch of 400,000 GPUs is on the verge of being deployed, alluding to the escalating competition in the realm of computing power.
