OpenAI Launches GPT-6 Astra, Signaling Arrival of AGI Era

OpenAI has officially launched GPT-6 Astra, its latest frontier artificial intelligence model, which the company describes as its most intelligent and aligned system to date. The model is designed to transition from simple question-answering to performing complex, multi-step tasks across professional domains, including software engineering, cybersecurity, scientific analysis, and automated computer use. OpenAI President Greg Brockman announced the release during a press briefing, framing the development as a significant milestone in the evolution of artificial intelligence.

While OpenAI’s official documentation emphasizes the model's technical capabilities and safety protocols, the launch has reignited discussions regarding the arrival of Artificial General Intelligence (AGI). During the briefing, Brockman stated, "For me personally, I do think we're there," and added, "Welcome to the AGI era." The release follows a period of heightened scrutiny regarding AI safety, particularly after security incidents involving models tested at the AI platform Hugging Face earlier this summer.

GPT-6 Astra represents a shift toward autonomous agentic behavior, with a primary focus on computer use. The model is capable of executing tasks such as filling out online forms, organizing calendars, conducting web research, and managing CRM records. OpenAI claims that Astra can significantly reduce the time required for complex workflows; for instance, the company stated that the model could potentially cut the time needed for apartment hunting from six hours to under 10 minutes.

Beyond administrative tasks, Astra demonstrates advanced proficiency in technical fields. It is designed to operate across codebases, test applications, and manage long-form coding workflows with minimal human intervention. In internal evaluations, the model showed significant performance gains over its predecessors, scoring 72.6% on the OSWorld 2.0 benchmark compared to 65.7% for the previous GPT-5.6 Sol model. Furthermore, the model achieved a 57.9% score on Terminal-Bench 4.0, marking a substantial improvement in its ability to handle complex software engineering challenges.

Cybersecurity is a central pillar of the Astra release. OpenAI reports that the model has reached a "Critical" cybersecurity capability threshold, meaning it possesses the ability to identify and develop functional exploits for vulnerabilities in hardened systems. During internal testing, Astra successfully discovered two previously unknown vulnerabilities, which were subsequently disclosed to the relevant maintainers.

Due to the dual-use nature of these capabilities, OpenAI is implementing a controlled rollout. Access to the most advanced cybersecurity features is currently restricted to trusted defenders through the company's Daybreak Blue program. OpenAI Chief Scientist Jakub Pachocki emphasized that the company remains cautious about the risks associated with such powerful models. "A model can become very good at achieving a goal, and it can still act in ways that go against what the person intended," Pachocki noted, adding that the company is prepared to slow down development if safety confidence is insufficient.

In addition to its coding and cybersecurity prowess, Astra has demonstrated high performance in scientific and mathematical reasoning. The model recorded a 99.9% score on the ARC-AGI-3 benchmark and 97.6% on FrontierMath Tier 4. These results suggest that Astra is capable of performing high-level professional tasks, including the creation of complex documents, presentations, and data analyses. The model also shows improved memory retention, allowing it to retrieve information from earlier parts of a conversation without relying solely on explicit notes.

OpenAI’s documentation highlights that Astra is state-of-the-art on several specialized benchmarks, including Terminal-Bench Science 0.1 and HealthBench Pro. By integrating these capabilities with web search, file search, and code execution tools, OpenAI is positioning Astra as an end-to-end solution for professional work. The company believes this model will enable a new generation of scientific discovery and entrepreneurship, despite the inherent uncertainties that accompany the deployment of such advanced systems.

OpenAI has begun a limited rollout of GPT-6 Astra to select organizations, with plans to expand access to ChatGPT Plus, Pro, Business, and Enterprise users in the coming days. The model will also be available via the OpenAI API and AWS. CEO Sam Altman acknowledged the anticipation surrounding the release, stating on social media, "We are working towards getting Astra in everyone's hands as quickly as we can; I know it is frustrating and I appreciate the patience."

The launch day was marked by widespread service disruptions across the AI sector. Users reported significant issues with ChatGPT, including problems with logins, file uploads, voice mode, and image generation. Similar outages were observed on the Claude and Grok platforms. While the exact cause of the simultaneous disruptions remains unclear, OpenAI has not confirmed any link between the Astra rollout and the ChatGPT outage. As the company moves forward, it continues to face external pressure, including an ongoing investigation by the U.S. state of Alabama regarding previous security breaches.