OpenAI's GPT-5.4 Can Now Use Your Computer For You

OpenAI has released GPT-5.4, a new model that moves the technology closer to the long-promised future of hands-free digital assistants. The update’s most significant feature is native computer control, enabling the AI to operate software, manage documents, and navigate applications directly on a user’s machine.

This represents a concrete shift toward autonomous agents—AI systems designed to execute multi-step tasks without constant supervision. The model can generate code to control a computer and issue keyboard and mouse commands based on screen analysis. OpenAI also notes improved web browsing and more precise use of external tools and APIs.

For reasoning, a specialized version called GPT-5.4 Thinking is being integrated into ChatGPT. It shows its work process for complex questions, allowing users to adjust their request mid-stream. The company states this model is better at synthesizing information from various sources and is 33% less likely to state falsehoods than its GPT-5.2 predecessor.

GPT-5.4 is available now through OpenAI’s API, its Codex coding tool, and in ChatGPT. The Thinking model is accessible for Plus, Team, and Pro subscribers, with a high-performance 'Pro' tier offered for Enterprise and education users. The feature is live on web and Android, with iOS support expected soon.