OpenAI took the wraps off GPT-6, with this release branded GPT-6 Astra. Greg Brockman, the company's president, said he anticipates the model will ultimately be regarded as an early stage of artificial general intelligence, and he defined AGI as systems that beat human performance across a broad array of tasks. He added, "I think that there's still lots of improvements to be made, but there is something significant here that I think is qualitatively improved." In his view, it signals "a real shift in what kind of work people can delegate to AI and how it can empower them," and he would "leave it up to the reader to decide for themselves if this qualifies for them."
OpenAI said Thursday it has made GPT-6 Astra available to business users through its Daybreak track, a pathway that gives vetted testers access to the company's most capable models. A separate version with extra cybersecurity guardrails is planned for paid ChatGPT users. In both cases, OpenAI said safeguards are in place to prevent access to the system's most cutting-edge cybersecurity capabilities.
OpenAI and Anthropic are racing to build more advanced AI while balancing heightened safety concerns, particularly after multiple security episodes tied to their systems. One flashpoint was an unintentional intrusion executed by OpenAI models against Hugging Face in July, which stoked worries about autonomous AI agents and spurred renewed calls from tech and government leaders for limits on the technology.
According to earlier statements from OpenAI, Astra has crossed a "critical cybersecurity threshold," meaning the system is capable of discovering and generating zero-day exploits without human help. The company increased the model's guardrails to reduce the risk of misuse, especially in cybersecurity contexts. The change followed an August call to suspend certain internal efforts on the model so tighter safeguards could be implemented, after concluding the system demonstrated strong aptitude on cybersecurity tasks.
When change is constant, steady investing still matters, so download the free Always Be Buying E-Book today
According to OpenAI, Astra is notably strong at operating a computer for a user, tackling a mix of work tasks and sticking with longer, more complex jobs than earlier models. The rollout pattern is telling: businesses in a controlled program get it first, and a broader consumer version comes with added guardrails. If you rely on AI inside the tools you use every day, expect capability upgrades to show up unevenly and with stricter limits around anything security related. That can influence which apps feel meaningfully smarter first, how fast productivity gains arrive in your workflow, and where the real value lands in the near term.
