OpenAI has unveiled its latest artificial intelligence system, describing it as the world’s most advanced model and a potential early step towards artificial general intelligence (AGI).
The company said the system combines years of research and “big bets” across pre-training, reinforcement learning and alignment, with capabilities spanning computer use, web browsing, software engineering and scientific work.
Cybersecurity is a major focus. OpenAI said the model demonstrated unprecedented cybersecurity and agent coding capabilities, prompting the company to halt testing last month.
The system began rolling out on 4 September to a limited group of organisations. It will become available to ChatGPT Plus, Pro, Business and Enterprise users, as well as through the OpenAI API, Microsoft Azure and AWS Bedrock.
OpenAI president Greg Brockman, in comments cited by Bloomberg, said the latest release represented a significant leap in capability and could eventually be viewed as an early stage of AGI.
“I think that there’s lots of improvements to be made, but there is something significant here that I think is qualitatively improved,” Brockman said.
He added that the technology represented “a real shift in what kind of work people can delegate to AI and how it can empower them”.
The announcement comes as OpenAI seeks to regain the technical lead from rival Anthropic, with the two companies competing to develop increasingly advanced AI systems.
The AI industry broadly defines AGI as software capable of outperforming humans across a wide range of tasks.
