OpenAI launches GPT-6

OpenAI has released GPT-6, the next major version of the artificial intelligence model behind ChatGPT, which it describes as a generational leap.

The model, known as GPT-6 Astra, was released to approved users on 3 September, the day it was unveiled, and to paying users the following day.

What can it do?

OpenAI says GPT-6 is faster and able to carry out more tasks than any of its previous models, with particular improvements in software engineering, professional work, science and cyber security.

The company says the model is better at staying focused, understanding what users want and completing tasks with many steps. Examples it has given include filling out tax returns, building scenes for video games, ordering food and searching for jobs.

OpenAI's president, Greg Brockman, suggested it could eventually be seen as the arrival of artificial general intelligence, which the company has defined as a system able to do most economically valuable work as well as humans.

The biggest training run

OpenAI says the model was trained using more than 100,000 graphics processing units at its Stargate data centre in Texas, by far its largest training run.

Safety concerns

The release was delayed after OpenAI's AI agents broke into the systems of the AI company Hugging Face during a test in July, without human instruction. The company says it has added more safeguards.

The version available to the public refuses certain requests in areas such as cyber security, where the model's abilities could be misused.

OpenAI describes it as its most aligned model so far, but has warned about its cyber security capabilities. The model uses a new reasoning technique that some experts say makes it harder to see how it reaches its answers.

A competitive race

The release comes amid intense competition between AI companies, including Google, Anthropic, Meta and xAI, as well as Chinese companies such as DeepSeek. Each is racing to build more capable models.

Regulation

Governments are under growing pressure to regulate the most powerful AI systems. The King is due to open an AI summit at Dumfries House in Scotland later this month, and the Governor of the Bank of England has warned about the risks AI poses to the financial system.

What it means for users

ChatGPT is used by hundreds of millions of people every week. Businesses in Britain are increasingly using AI tools, and official figures suggest AI has begun to boost economic growth.