Exactly seven weeks after release ChatGPT 5.4OpenAI is dropping a new family of models. ChatGPT 5.5 is available now for paid ChatGPT and Codex users, as well as the virtual and pro versions, with plans to bring it to the API soon. These models are designed for work — specifically, coding, computing and research.
OpenAI President Greg Brockman said the GPT-5.5 is more intuitive than previous models and can do more with less human guidance.
“It can look at a vague problem and figure out what needs to happen next,” Brockman told reporters at a briefing. “It feels like it’s laying the groundwork for how we’re going to do computing going forward, or how agent computing at scale is going to work.”
This is a standard model, so anyone can use it. But it will probably be more useful for people doing research or other intensive tasks, like coding. It has the power of an agent, meaning it can complete tasks independently. It scored higher than the GPT-5.4 in benchmarks that measure the model’s ability to run applications across your computer and solve math problems.
OpenAI and other tech companies are trying to build models that can act as a real digital assistant that can manage notifications and oversee projects across your computer, not just a single system. GPT 5.5 is the company’s biggest turn in this yet.
The company says GPT-5.5 has its strongest cybersecurity protections to date, and when it launches, it will be even more robust in enabling applications regarding cybersecurity. This is due to the growing concern that cyber-powered AI models could find weaknesses in our existing cyber infrastructure. That is why both are Anthropic A preview of the Claude Mythos and OpenAI’s 5.4-Cyber released to developers to iron out issues before a potential public release.
AI in research
At OpenAI, research is a two-way street. GPT 5.5 is part of the company’s super app foundation aims to build without the Codex. Research on AI is also increasingly being done with the help of AI tools; Brockman said the company used 5.5 (and Codex) to help build it during development.
AI has changed the way research is done in the past few years, and the field is not immune to the concerns of other industries AI is likely to automatically put people out of work in the future. Although GPT-5.5 is OpenAI’s most capable research model yet, the company’s senior scientists and researchers are not worried about being replaced by AI.
Mark Chen, chief research officer of OpenAI, talked about the pursuit of the near-term goal of having humans become the “editors” of AI-assisted research. Mia Glaese, vice president of research, said she hopes the AI models’ capabilities mean more research can be done, “raising the bar of what’s worth building.”
“I feel more productive because the challenge is changing from getting implementation details, low-level abstractions, to high-level goals,” said Jakub Pachocki, chief scientist of OpenAI. “It allows you to make progress much faster and use your focus, your energy to find out, ‘What are the important things?’…There’s no need to do things for the sake of them.”
