OpenAI announced Thursday the launch of GPT-6 Astra, a new model it claims is the current best at controlling computers, web browsers, writing code, and solving complex mathematics.
Availability for paying customers begins in the coming days. Enterprise clients enrolled in the Daybreak early access programme get first access, followed by subscribers to ChatGPT Plus, Pro, and Business plans. The company did not confirm whether free users of ChatGPT will receive the update.
During a press briefing, OpenAI executives expressed strong confidence in the new system. Greg Brockman, president and co-founder, stated he believes the world has entered the era of artificial general intelligence, or systems generally smarter than humans.
“It’s not unreasonable to feel that we are now in the AGI era,” Brockman told reporters. “I think that if we fast forward a couple of years, when we look back and say, ‘When was it really that AGI was created?’ I think it’s going to be about this time, and I think it might be about this model.”
The release comes at a difficult time for the company. OpenAI aims to advance its technology and drive sales before a planned initial public offering while maintaining safety and security standards. It also faces stiff competition from Anthropic, its main rival, which is preparing for its own initial public offering.
In a blog post, OpenAI describes GPT-6 Astra as the “world’s best computer use model.” The system can navigate computers and web browsers on behalf of humans better than previous versions. Historically, AI companies have struggled to build models that reliably browse websites, fill out forms, and manipulate software tools quickly and accurately enough to replace humans. In tests, GPT-6 Astra booked DMV appointments, searched for job listings, and hunted apartments faster than the average person.
The company also claims this is the safest and most aligned AI model OpenAI has released. Jakub Pachocki, chief scientist, told reporters that the firm is monitoring GPT-6 Astra’s chain-of-thought, the internal scratchpad the model uses to work through problems, to ensure safe deployment. Pachocki noted that preventing advanced AI models from causing unintended harm is becoming an increasing challenge and may act as a bottleneck to further progress.
“We kind of take this visibility for granted, and we are seeing that as model capabilities are increasing, monitorability is getting more challenging,” Pachocki said. “We think confidence in monitoring may constrain further development, because we would not accept degradation in our ability to monitor model alignment beyond a certain level. We would withhold scaling until we can regain enough confidence.”
What it means
For people making things, the ability to interact with software faster than a human suggests a shift in how work gets done. Tasks that previously required manual clicking or form-filling could now be handled by the AI. However, the warning from OpenAI’s scientists suggests that this speed comes with a catch. If the ability to verify what the AI is thinking degrades, the company may stop expanding the model’s power. This creates a potential ceiling on how much these tools can improve, as safety checks become harder to maintain.
