The new frontier model brings advances in computer use, coding, cybersecurity, science and complex professional workflows
OpenAI has introduced GPT-6 Astra, its latest frontier model, describing it as its most intelligent and aligned model to date. The model began rolling out on September 3 to a limited set of organisations, with broader access planned for ChatGPT Plus, Pro, Business and Enterprise users, as well as developers through the OpenAI API, Microsoft Azure and Amazon Bedrock.
Astra is designed to handle complex, end-to-end tasks spanning reasoning, coding, computer use, research and professional work. OpenAI says the model represents a significant step forward in its ability to carry out multi-step workflows, including working with browsers and professional software, creating documents and presentations, and adapting when users introduce new requirements during a task.
The model has also been positioned as a major advancement in computer-use capabilities. Astra can handle tasks such as completing online forms, updating CRM records, organising calendars, conducting online research and working within document editors. OpenAI says it delivers improvements in speed, accuracy and judgment when handling demanding computer-based workflows.
Beyond general productivity, Astra has been developed with stronger capabilities in areas including software engineering, science and cybersecurity. OpenAI says the model achieves state-of-the-art results across several evaluations, including computer-use and cybersecurity benchmarks.
OpenAI has also placed significant emphasis on Astra’s alignment and safety capabilities. The company says the model was developed to better understand user intent, respect task boundaries and communicate more transparently when handling ambiguous instructions. Additional monitoring has been introduced to identify potential cases where an agent may not have interpreted instructions correctly, with systems able to pause or stop a conversation for review when necessary.
Cybersecurity is another area where Astra marks a significant milestone for OpenAI. According to the company’s safety assessment, GPT-6 Astra is its first model to reach the Critical level of cybersecurity capability under its Preparedness Framework. OpenAI says the model can, with the right tools and access, identify previously unknown vulnerabilities and develop ways to exploit weaknesses in highly protected systems. This has led the company to introduce additional safeguards around the model’s deployment and use.
The launch follows OpenAI’s recent focus on strengthening model security and alignment following the Hugging Face security incident. The company has outlined additional measures including stronger monitoring and security controls as it develops increasingly capable models.
For developers, GPT-6 Astra is available through the API under the model name gpt-6-astra. OpenAI lists support for reasoning, coding, computer use, web search, file search, image generation, code interpreter and other tools. The model has a 1.05-million-token context window and supports up to 128,000 output tokens.
GPT-6 Astra is being rolled out in phases, with access initially limited to selected organisations before expanding to ChatGPT Plus, Pro, Business and Enterprise users and API customers.
The launch positions Astra as OpenAI’s latest attempt to move AI systems beyond generating answers towards completing entire workflows on a user’s behalf – combining reasoning, tool use, computer interaction and specialised capabilities within a single model.