GPT-6 Astra: OpenAI Unveils a New AI Generation With Major Leap in Reasoning, Computer Use and Cybersecurity
OpenAI has introduced GPT-6 Astra, its latest flagship artificial-intelligence model, describing it as a major step forward in reasoning, computer use, software engineering, scientific research and professional work. The model was officially released on September 3, 2026, with broader access being rolled out in phases.
Astra is designed to move beyond simply answering questions. OpenAI says the model can perform complex, multi-step tasks across browsers, software applications and professional tools, including filling forms, updating records, conducting research, analysing data and creating documents, presentations and spreadsheets.
One of the biggest changes is its ability to operate computers more effectively. On OpenAI’s cited OSWorld 2.0 evaluation, Astra scored 72.6%, compared with 65.7% for GPT-5.6 Sol, while completing tasks in substantially less simulated time. The company says this could make AI agents considerably more practical for real-world office and technical workflows.
The model also posts striking results on advanced reasoning benchmarks. OpenAI reports a 98% score on FrontierMath Tier 4 and 99.9% on ARC-AGI-3, alongside a 100% result on its cited ExploitBench evaluation. These results are intended to demonstrate progress on difficult mathematical reasoning, novel problem-solving and cybersecurity tasks.
Cybersecurity, however, is also one of the most sensitive aspects of Astra. OpenAI says the model has reached the “Critical” level of cybersecurity capability under its Preparedness Framework, meaning that with appropriate tools and access it can potentially identify previously unknown vulnerabilities and develop sophisticated exploits.
That capability has prompted stronger safeguards. OpenAI says it has introduced additional protections, including tighter isolation of systems, checkpoint encryption, broader monitoring and alignment evaluations before deployment. The company acknowledges that the greater defensive potential of the model comes alongside greater risks if the technology is misused.
Astra is also being positioned as a model for professional industries rather than only a consumer chatbot. OpenAI says it can work with complex templates and produce polished business documents, presentations, spreadsheets and analyses while maintaining the structure and style required by organisations.
The financial sector is already emerging as an important early application. OpenAI recently launched ChatGPT for Financial Services, incorporating GPT-6 Astra for investment banking and equity-research workflows. The system can connect with financial-data providers and assist with research, financial modelling and preparation of client materials.
The model’s capabilities also come with a significant change in the economics of AI. OpenAI lists API pricing at $10 per million input tokens and $50 per million output tokens, although the company argues that Astra’s ability to complete difficult tasks with fewer tokens can reduce the overall cost of some workflows.
Astra has a context window of 1.05 million tokens and can produce up to 128,000 output tokens, according to OpenAI’s API documentation. It supports multiple reasoning levels, allowing developers to trade off speed, computational effort and performance depending on the task.
The release nevertheless arrives amid growing debate over the safety of increasingly capable AI systems. OpenAI’s own safety documentation makes clear that Astra represents a substantial increase in capability, particularly in cybersecurity, while also stressing that stronger safeguards are required as models become more autonomous.
For ordinary users, the significance of GPT-6 Astra may ultimately be less about benchmark scores and more about delegation. Instead of merely asking an AI to explain how something should be done, users can increasingly ask it to carry out the work across multiple steps, applications and tools.
That shift could make Astra an important milestone in the transition from AI as a conversational assistant to AI as an active digital worker—while simultaneously making questions about security, oversight, accountability and human control more urgent than ever.
