A Major Leap Beyond Chatbots
OpenAI has launched GPT-6 Astra, its latest flagship artificial intelligence model, describing it as the company’s most intelligent and aligned system yet. Released on September 3, 2026, Astra is designed to move AI beyond simply answering questions and toward independently completing complex, multi-step tasks.
The model is initially being rolled out to a limited group of organisations, with availability expanding to ChatGPT Plus, Pro, Business and Enterprise users, as well as developers through the OpenAI API and major cloud platforms.
AI That Can Operate a Computer
One of Astra’s biggest upgrades is its ability to interact directly with computers.
The model can fill out online forms, update customer records, organise calendars, conduct web research and prepare information in documents or emails. It can also work with spreadsheets, analyse scientific data, create websites and perform software testing.
This changes the role of an AI assistant. Instead of telling a user how to complete a task, Astra is designed to actually carry out much of the work.
OpenAI says Astra completes computer-based tasks significantly faster than its predecessor while achieving higher performance on several computer-use evaluations.
Stronger Coding and Professional Skills
Software engineering is another major focus. OpenAI describes Astra as its strongest model for coding to date, with improvements in understanding codebases, debugging and completing complex development workflows.
The model has also been trained for professional work involving documents, presentations, spreadsheets and data analysis. It can follow existing templates, understand formatting requirements and produce outputs intended to be immediately usable.
For businesses, this could make AI agents more practical for routine knowledge work, potentially automating tasks that previously required employees to move information manually between different applications.
Big Gains in Science and Cybersecurity
Astra also claims major advances in scientific reasoning and cybersecurity. OpenAI reports extremely high scores on several frontier benchmarks, including a 100% result on its ExploitBench cybersecurity evaluation.
The cybersecurity capabilities are being handled particularly cautiously because of the potential for misuse. Access to the most powerful security features is initially restricted to trusted organisations and defensive applications.
OpenAI has also emphasised alignment and control, saying Astra is better at understanding user intent and staying within authorised boundaries.
A Step Toward AGI?
The launch is being presented as more than another model upgrade. OpenAI sees Astra as a potential step toward artificial general intelligence, particularly because it combines reasoning with the ability to take actions in real-world software environments.
That distinction could prove important. The next phase of AI competition may not be about which chatbot gives the best answer, but which AI system can reliably complete an entire workflow with minimal human intervention.
Astra therefore represents a shift from AI that responds to AI that acts—although questions around safety, reliability, access and human oversight will become increasingly important as these capabilities expand.
Summary
OpenAI has launched GPT-6 Astra, a new AI model focused on computer use, coding, science, cybersecurity and professional work. Unlike conventional chatbots, Astra can perform multi-step tasks across software applications, including filling forms, updating records, conducting research and creating business documents. OpenAI says the model represents a major advance toward more capable AI agents and potentially the broader goal of AGI.