OpenAI Launches GPT-6 Astra to ChatGPT Subscribers Amid Cybersecurity Concerns

OpenAI's latest AI model, GPT-6 Astra, officially launched on September 3, 2026, starting with a select group of testers before expanding to paid ChatGPT subscribers across various tiers including Plus, Pro, Business, and Enterprise. Users reported that Astra appeared in their accounts shortly after the initial deployment, highlighting the company's commitment to a smooth rollout.

OpenAI president Greg Brockman described Astra as a "generational leap," signaling the start of what the company considers the AGI era. The model boasts impressive performance metrics, completing tasks approximately 47% faster than its predecessor, GPT-5.6 Sol. On the OSWorld 2.0 benchmark, Astra achieved a score of 72.6% in 40 minutes, surpassing Sol's 65.7% in 75 minutes. Furthermore, Astra scored 98% on FrontierMath Tier 4 and a perfect 100% on ExploitBench, indicating its ability to autonomously detect and exploit unknown vulnerabilities in secure systems. This capability earned Astra OpenAI's highest internal classification of "Critical" on its Preparedness Framework for cybersecurity.

Due to the advanced features of Astra, OpenAI has initially restricted access to some of its most sophisticated cybersecurity functionalities. The rollout faced challenges, as many ChatGPT Plus and Pro users experienced access issues, prompting CEO Sam Altman to issue a public apology. He acknowledged the confusion among subscribers who expected immediate access to the new model, and OpenAI offered banked resets as compensation for the disruptions.

Despite the rollout challenges, Astra's introduction represents a significant upgrade for businesses already utilizing ChatGPT, providing them with a more capable system at no additional cost. Brockman's characterization of Astra as the beginning of the AGI era is ambitious, yet the model's benchmarks lend credibility to this claim.