What happened
On September 6, Nvidia CEO Jensen Huang posted on social media that OpenAI's GPT-6 Astra model was trained using over 100,000 Nvidia Grace Blackwell NVLink 72 clusters.
Huang congratulated the OpenAI team and stated plainly that 'artificial general intelligence (AGI) has arrived.'
OpenAI released GPT-6 Astra on September 3, calling it a 'generational leap' with advanced capabilities in computer operation, software engineering, cybersecurity, and scientific research.
Why it matters
This declaration from a leading industry figure signals a major milestone in AI development, potentially reshaping expectations for AI's role in various sectors.
The scale of the training infrastructure highlights the immense computational resources required for cutting-edge AI models, underscoring the importance of advanced hardware like NVLink 72.
Key facts
GPT-6 Astra was trained on over 100,000 Nvidia Grace Blackwell NVLink 72 clusters.
NVLink 72 interconnects 72 GPUs to function as a single 'giant GPU.'
OpenAI released GPT-6 Astra on September 3, describing it as a 'generational leap.'
What to watch next
Watch for further validation or debate within the AI community about whether GPT-6 Astra truly meets the criteria for AGI.
Monitor how this milestone influences AI regulation, investment, and the competitive landscape among tech companies.