OpenAI Launches GPT‑5.4, an AI That Can Operate Computers

San Francisco: OpenAI has introduced GPT‑5.4, a powerful new AI model that, unlike earlier models, can use a computer on its own, opening apps, clicking buttons, typing, and even understanding what is happening on the screen. 

The new model comes in two versions, namely GPT‑5.4 Thinking and GPT‑5.4 Pro. Both versions are programmed to make complex office work easy by handling tasks across spreadsheets, documents, presentations, and software tools. The model can read screenshots, send keyboard and mouse commands, and follow multi‑step instructions much like a human assistant. 

GPT‑5.4 also supports a massive 1 million‑token context window, that is to say, it can process very large files, entire codebases or long documents, without losing track. It combines major improvements in reasoning, coding, and long‑running task execution, making it much more reliable for professional use, especially in AI-powered spreadsheet tools. According to OpenAI, it is 33% less likely to produce incorrect information compared to earlier versions. 

The update includes 6 major boosts, such as better coding abilities, improved image understanding, and stronger performance on tasks that involve multiple steps. It also performs well in areas like legal analysis, financial modelling, and creating slide decks, helping users complete detailed work in a quick and efficient manner. 

With GPT‑5.4, OpenAI moves beyond AI that only answers questions. For this new model can now take action and boost your workplace productivity like never before.