OpenAI Launches GPT-6 Astra AI Model, Declaring Beginning of 'AGI Era'
San Francisco-based AI research organization OpenAI has officially unveiled its next-generation flagship model, named GPT-6 Astra, marking what the company describes as a monumental leap forward in artificial intelligence capabilities. Released in limited preview on September 3, 2026, the new model succeeds GPT-5.6 Sol and is engineered to act as an autonomous computer operator rather than a traditional conversational chatbot.
GPT-6 Astra introduces major technical advances across complex reasoning, software engineering, scientific research, and direct computer and browser navigation. According to benchmark evaluations released by OpenAI, Astra achieved unprecedented scores on rigorous testing suites, including 97.6% on FrontierMath Tier 4 and 99.9% on ARC-AGI-3 environments. Crucially, it is the first model to cross the 'Critical' cybersecurity threshold under OpenAI’s Preparedness Framework, demonstrating the capability to discover and exploit zero-day software vulnerabilities independently without human intervention.
Following the announcement, OpenAI President Greg Brockman stated that the technology industry has entered a new era of Artificial General Intelligence (AGI), pointing to the model's ability to execute complex end-to-end workflows such as preparing tax returns, building complete websites, and solving open mathematical problems in minutes. However, because of Astra's potent cybersecurity capabilities, OpenAI has implemented strict safety safeguards. The model's full capabilities will initially be restricted to vetted enterprise partners under the Daybreak safety initiative before a broader rollout to paid ChatGPT Plus, Pro, and enterprise subscribers over the coming days.
