OpenAI Launches GPT-6 Astra With Advanced Computer Use and Coding Abilities

OpenAI has launched GPT-6 Astra, its latest frontier artificial intelligence model, with significantly enhanced capabilities in computer use, coding, cybersecurity and complex, long-running tasks. The new model is designed to move beyond simply answering questions and instead carry out multi-step tasks with greater autonomy.

According to OpenAI, Astra can operate computers in a manner similar to a human user. It can browse the internet, complete online forms, update customer relationship management systems, manage calendars and conduct research. The model can also create and edit documents, spreadsheets and presentations while following existing templates and visual styles.

The model’s expanded capabilities are being viewed as a potential step towards artificial general intelligence (AGI), where AI systems could perform a broad range of intellectual tasks rather than being limited to specific functions. Its ability to independently plan and execute several connected actions marks a shift from conventional prompt-based AI assistants. GPT-6 Astra also features a context window of 1.05 million tokens and can produce up to 128,000 output tokens. This allows it to work with large quantities of information and sustain complex tasks over longer periods.

Coding is another major area of improvement. Astra is designed to assist developers with complex programming tasks, including writing and testing software. Its enhanced reasoning and ability to work with computers could allow it to handle larger portions of software-development workflows with less human intervention.

However, the model’s greater capabilities have also raised concerns about cybersecurity. OpenAI has said Astra can identify and exploit security vulnerabilities with limited human guidance, prompting the company to introduce stronger safety measures before and around its deployment. The cybersecurity capabilities have become particularly significant as increasingly autonomous AI systems are being tested in real-world environments. While such technology could help security professionals discover vulnerabilities and strengthen digital systems, the same abilities could potentially be misused.