OpenAI launches GPT-6 Astra as company claims major leap towards AGI

OpenAI has officially launched GPT-6 Astra, its latest and most advanced artificial intelligence model, with the company describing the release as a potential turning point in the development of artificial general intelligence (AGI).

Announced on 3 September 2026, GPT-6 Astra is designed to move beyond simply answering questions and generating text, allowing it to carry out complex, multi-step tasks using computers, software and online services.

OpenAI says Astra is its most intelligent and aligned model to date, with significant improvements in reasoning, coding, computer use, scientific research and professional tasks.

Designed to use computers

One of the biggest changes with GPT-6 Astra is its ability to interact with computers in a way that more closely resembles a human user.

The model can work with applications through their graphical interfaces, allowing it to perform tasks such as filling in forms, updating CRM systems, working with spreadsheets, carrying out web research and operating engineering software.

This represents a significant change from traditional chatbots, where users normally have to take the information produced by an AI system and carry out the final steps themselves.

With Astra, OpenAI is instead aiming for an AI system capable of completing an entire workflow from beginning to end.

Major improvements for coding

Software development is another major focus of GPT-6 Astra.

OpenAI says the model is capable of handling complex software engineering tasks, including writing, testing and debugging code. It can also work with development environments and computer tools rather than simply producing snippets of code in a conversation.

The company believes this could make Astra particularly useful for developers working on large and complicated software projects.

Powerful cybersecurity capabilities

Perhaps one of the most controversial aspects of GPT-6 Astra is its cybersecurity capability.

OpenAI says Astra is the company’s first model to reach the “Critical” level of cybersecurity capability under its Preparedness Framework. The company has therefore introduced additional restrictions around some of its most powerful cybersecurity capabilities.

The model is capable of identifying serious security vulnerabilities and assisting with advanced defensive cybersecurity work.

However, OpenAI is also concerned that increasingly capable AI systems could potentially be misused by attackers. As a result, access to some of Astra’s strongest cybersecurity capabilities is initially being restricted to trusted defenders.

A step towards AGI?

OpenAI has positioned GPT-6 Astra as more than simply another incremental improvement to ChatGPT.

The company has described the release as potentially marking the beginning of what it calls the AGI era.

Astra is intended to perform increasingly complex tasks autonomously and combine reasoning, computer interaction and specialist knowledge. OpenAI believes these capabilities could eventually allow AI systems to carry out a significant proportion of economically valuable work.

However, the AGI claim is likely to remain controversial, with researchers and technology experts continuing to debate exactly what should qualify as artificial general intelligence.

Safety concerns remain

The launch also comes at a time when the AI industry’s approach to autonomous systems is facing increased scrutiny.

OpenAI has acknowledged that its most capable models can be increasingly difficult to monitor, particularly as they become better at planning and completing complicated tasks.

The company says it has introduced additional safeguards around GPT-6 Astra and is continuing to develop systems capable of detecting and stopping potentially dangerous behaviour.

The concerns demonstrate the difficult balance facing AI developers: making models increasingly capable while ensuring humans remain able to understand, control and safely deploy them.

Availability

GPT-6 Astra is being rolled out gradually rather than immediately becoming available to everyone.

The initial deployment is focused on selected trusted and enterprise users, with broader access to paid ChatGPT customers and developers expected to follow.

OpenAI has also made Astra available through its developer platform, allowing businesses and developers to integrate the model into their own applications.

The next chapter for AI

GPT-6 Astra represents a significant change in OpenAI’s vision for artificial intelligence.

Rather than simply providing answers, the company wants AI to become capable of taking action, operating software and completing complicated tasks on behalf of its users.

If Astra performs as OpenAI claims, the launch could mark an important step towards increasingly autonomous AI assistants.

But with greater capability comes greater risk. How successfully OpenAI can keep such powerful systems safe while making them widely available may ultimately prove just as important as the technology itself.

For now, GPT-6 Astra represents OpenAI’s clearest attempt yet to move AI from a tool that responds to commands towards an intelligent system capable of actually getting the job done.