“AI Disruption” Publication 10,500 Subscriptions 20% Discount Offer Link.
GPT-6 Astra arrives without warning.
Just now, OpenAI officially positioned it as “the world’s most intelligent model, and the one most aligned with human intent.”
OpenAI said Astra brings together years of research in pretraining, reinforcement learning, and model alignment, reaching new capability levels across computer use, web browsing, software engineering, cybersecurity, scientific research, and professional work.
Compared with previous models that mainly focused on chat, writing, and Q&A, Astra is more like an AI agent that can actually get work done: it can call tools, operate software, browse the web, and complete complex tasks independently. This release is also OpenAI’s largest model upgrade in recent years.
According to OpenAI research lead Aidan Clark, Astra is one of OpenAI’s largest training projects to date, and the first next-generation model to make extensive use of earlier models for supervision during training. It is also the first model pretrained with more than 100,000 GPUs at the U.S. Stargate campus in Texas. OpenAI president Greg Brockman went even further: years from now, he said, people will look back on today as the starting point of the AGI era.
