OpenAI Unveils GPT-5.4, AI That Can Operate Your Computer Like a Human
GPT-5.4 can now operate a computer independently, perform complex tasks, and improve workplace productivity across multiple applications.
OpenAI is once again pushing the boundaries of artificial intelligence with its latest release, GPT-5.4, a model designed to handle professional workloads and, for the first time, operate a computer on behalf of users. Moving beyond simple text generation, this new AI can navigate spreadsheets, documents, and presentations, transforming the way digital tasks are completed.
“AI tools are gradually moving beyond answering questions or generating text. Technology companies are now trying to build systems that can carry out tasks on a computer in the same way a person would,” the company noted in its announcement.
GPT-5.4 combines enhancements in reasoning, coding, and productivity workflows to provide a more capable AI assistant. The model’s standout feature is its native computer-use capability, allowing it to execute keyboard and mouse commands and respond to on-screen information. This enables GPT-5.4 to identify what needs to be done and take actions across multiple software applications.
According to OpenAI, “This allows the AI system to identify what needs to be done next and perform tasks across different software tools.” Essentially, GPT-5.4 can act as a digital assistant that completes tasks autonomously, from drafting presentations to managing spreadsheets, saving users valuable time.
OpenAI has also introduced GPT-5.4 Thinking, a reasoning-focused variant being rolled out inside ChatGPT. This version is engineered for multi-step problem-solving and provides users with a clear outline of its approach to complex queries. Users can even adjust their requests while GPT-5.4 Thinking is generating a response, eliminating the need to restart the process.
“This makes it easier to guide the model toward the exact outcome you want without starting over or requiring multiple additional turns,” OpenAI explained.
The new model also shows notable improvements in accuracy and information synthesis. OpenAI claims GPT-5.4 is the company’s “most factual model yet,” with individual claims 33% less likely to be incorrect compared to GPT-5.2. It can also gather information from multiple sources more effectively, making it useful for tasks that require careful research or cross-referencing.
Moreover, GPT-5.4 is designed to interact seamlessly with web browsers and external tools, calling APIs efficiently and completing multi-step tasks with greater precision. Developers will have access to the model through OpenAI’s API and the coding tool Codex, while enterprise and educational users can benefit from GPT-5.4 Pro for high-performance workflows.
Currently, GPT-5.4 Thinking is available on the ChatGPT web app and Android, with iOS support expected soon. OpenAI’s latest release signals a major shift in AI development, emphasizing autonomous digital assistants capable of performing real work across software environments rather than just generating text responses.
As AI continues to evolve, GPT-5.4 could redefine workplace efficiency, enabling users to delegate complex, multi-step tasks to intelligent systems that think, act, and adapt in real time.
