OpenAI says its new flagship model can handle complex computer tasks, coding, research and multi-step professional workflows.
OpenAI has launched GPT-6 Astra, its latest flagship AI model, on September 3, 2026. The company says Astra is designed to handle complex tasks across computer use, coding, research, cybersecurity, science and professional work.
The new model is being introduced with a focus on AI agents that can complete multi-step workflows. OpenAI says Astra can work with computers and professional software while producing documents, spreadsheets and presentations based on user instructions and existing templates.
What is GPT-6 Astra?
GPT-6 Astra is OpenAI’s latest high-capability model for complex end-to-end work. The company describes it as its most intelligent and aligned model, with improvements across reasoning, computer use, software engineering and scientific tasks.
A key feature is its ability to carry out several connected steps instead of only responding to individual prompts. This makes the model more suitable for tasks where users need an AI system to work through an entire process.
OpenAI says Astra can navigate computers and browsers, work with specialist software and create finished documents, spreadsheets and presentations. It is also designed to follow existing templates and adapt when requirements change.
Focus on autonomous AI agents
Astra is part of OpenAI’s wider push towards agentic AI. These systems are designed to perform tasks through multiple steps with less need for continuous user input.
According to OpenAI, Astra can combine reasoning with computer use to complete professional workflows. The model can also work with software tools to inspect data, make changes and check results.
For businesses and professionals, this could make AI useful for more than simple question-and-answer tasks. However, OpenAI’s deployment approach includes additional safeguards for workflows where the model can take actions on a computer.
Stronger performance in coding and research
OpenAI says GPT-6 Astra delivers state-of-the-art results across several computer-use, coding, science and professional benchmarks.
The company reports a 57.9% score on Terminal-Bench 4.0, compared with 37.3% for GPT-5.6 Sol. On GPQA Diamond, a benchmark for advanced scientific reasoning, Astra achieved a 96% score in the comparison published by OpenAI.
The model is also designed to support scientific work. OpenAI says Astra can use scientific software to inspect data and help researchers assess results and decide what to investigate next.
Cybersecurity becomes a major focus
Cybersecurity is one of the most important areas in the Astra launch. OpenAI says the model has reached the Critical level for cybersecurity capability under its Preparedness Framework.
The company says Astra can identify security weaknesses and, with the right tools and access, develop exploits against well-protected systems without step-by-step human guidance. OpenAI has therefore introduced stronger safeguards around the model’s deployment.
The company says the safeguards include stronger protection against jailbreaks, monitoring systems and controls designed to detect potentially unsafe or unauthorised activity.
OpenAI is also using its Daybreak programme to gradually expand access to advanced cybersecurity capabilities for authorised defensive work. The programme is intended to help security professionals find vulnerabilities, validate weaknesses and develop patches while maintaining controls against misuse.
GPT-6 Astra availability
OpenAI says GPT-6 Astra is initially rolling out to a limited group of organisations. Wider access is planned for ChatGPT Plus, Pro, Business and Enterprise users over the following days.
The model is also being made available through the OpenAI API and AWS. Enterprise administrators can control access to Astra within their workspaces.
For developers, OpenAI lists the model under the API identifier gpt-6-astra. The company says it supports tools including web search, file search, image generation, code interpreter, computer use and hosted shell.
What the launch means for users
GPT-6 Astra represents a shift towards AI systems that can complete larger pieces of work rather than simply generate individual responses.
Its combination of reasoning, computer use, coding and professional workflows could make it useful for developers, researchers and businesses. At the same time, OpenAI’s staged rollout shows the company is treating the model’s stronger capabilities, particularly in cybersecurity, as a significant safety issue.
OpenAI will continue expanding access to GPT-6 Astra in phases while maintaining additional safeguards for higher-risk uses.
For the lates updates, follow CKSOR News