OpenAI CEO Sam Altman during a media tour of the Stargate AI Data Center in Abilene, Texas, September 23, 2025.
Bloomberg | Bloomberg | Getty Images
OpenAI on Thursday announced its latest artificial intelligence model, GPT-5.5. The company says this model is excellent for coding, using computers, and pursuing deeper research skills.
This release comes less than two months after OpenAI released GPT 5.4, the latest sign of the breakneck pace of development driving the AI sector.
“What’s really special about this model is how much it can do with so little guidance,” OpenAI President Greg Brockman said at a press conference Thursday. “You can look at unclear problems and understand what needs to happen next. To me, this feels like it’s laying the foundation for how we will use computers and how we will do computer work in the future.”
OpenAI is racing to catch up with rivals like Google and Anthropic. Anthropic’s latest model, the Claude Mythos Preview, is captivating Wall Street.
According to OpenAI, GPT-5.5 excels at data analysis, writing and debugging code, working with software, online research, and creating documents and spreadsheets. The company added that the model does not exceed the “critical” cybersecurity risk threshold, which could result in “unprecedented new paths to serious harm,” but meets the criteria for a “high” risk classification, which could “amplify existing paths to serious harm.”
“GPT-5.5 has undergone extensive third-party safeguard testing and cyber and bio (risk) red teaming, and we have been iterating on cyber safeguards for several months as we further strengthen our cyber response model,” Mia Grace, OpenAI’s vice president of research, said in a Thursday briefing.
Cybersecurity risks posed by AI have been a top concern for technology executives and government officials since Anthropic introduced its Mythos model earlier this month. The company decided to limit the deployment of Mythos due to its ability to identify weaknesses and security flaws within the software.
GPT-5.5 will be rolled out to OpenAI’s paid subscribers, including Plus, Pro, Business, and Enterprise users, on ChatGPT and its coding assistant Codex on Thursday. The company said the model will be introduced to application programming interfaces “soon” but that their introduction requires “other safeguards.”
WATCH: Bradley Tusk: It’s hard to understand why investors would want OpenAI stock when it’s public
