OpenAI announced Thursday its next-generation artificial intelligence model, GPT-6 Astra. The company says the new AI system delivers state-of-the-art performance for computer and web browser use, software development, and advanced mathematical problem-solving.
OpenAI plans to make GPT-6 Astra available to paying customers in the coming days. Access will begin with enterprise customers participating in the Daybreak early access program before expanding to ChatGPT Plus, Pro, and Business subscribers. OpenAI has not confirmed whether GPT-6 Astra will be offered to free ChatGPT users.
During a press briefing, OpenAI executives expressed strong confidence in the capabilities of the company’s latest AI model. OpenAI President and co-founder Greg Brockman said he believes the world has entered the era of artificial general intelligence, or AGI—AI systems capable of outperforming humans across a broad range of tasks.
“It’s no wonder we feel like we’re in the era of AGI right now,” Brockman told reporters. “If you fast-forward a few years and look back and ask, ‘When was AGI actually created?’ I think it would probably be around this time—and probably around this model.”
The launch of GPT-6 Astra comes at a pivotal moment for OpenAI. The company is working to advance its AI technology, increase revenue, and prepare for planned products while maintaining strong safety and security standards. OpenAI also faces growing competition from Anthropic, another leading AI company reportedly preparing for a potential initial public offering.
In a blog post, OpenAI described GPT-6 Astra as “the world’s best computing model.” The company claims the model can use computers and web browsers on behalf of users more accurately and efficiently than earlier AI systems. AI developers have historically struggled to build models that can reliably navigate websites, complete online forms, and interact with software tools. According to OpenAI, GPT-6 Astra successfully made DMV reservations, searched job listings, and found apartment listings faster than the average person during testing.
OpenAI also said GPT-6 Astra is its most secure and carefully tuned AI model to date. Jakub Pachocki, OpenAI’s chief scientist, told reporters that the company monitors the model’s chain of thought—the internal reasoning process used to solve problems—to support safer deployment. However, Pachocki acknowledged that preventing unintended harm from increasingly capable AI systems is becoming more difficult and could limit future progress.
“We kind of take this visibility for granted, but we’ve found that as models become more capable, observability becomes more difficult,” Pachocki said. “We believe that confidence in monitoring may limit further development because we cannot accept that our ability to monitor the integrity of the model degrades beyond a certain level. We will hold off on expansion until we regain sufficient confidence.”
Source: www.wired.com