OpenAI today introduced GPT-5.4, the latest update to its flagship AI model, bringing improvements in reasoning, coding, and automation of real-world tasks. The model is rolled out across ChatGPT, the API and developer tools, with new variants targeting both everyday users and enterprise workloads.
One of the biggest changes is the model’s ability to interact more directly with computers. GPT-5.4 can interpret screenshots, operate browsers, and issue keyboard and mouse commands to complete tasks in various apps and services. This enables it to perform multi-step workflows that previously required human intervention. This represents a major step forward towards more autonomous AI agents.
The update also improves the model’s ability to examine complex questions. According to OpenAI, GPT-5.4 can search across multiple rounds of information gathering and combine results into clearer, more structured answers. The company claims the model is the most factual to date and reduces false claims by about 33 percent compared to GPT-5.2
A new “Think” mode for more difficult questions
In addition to the core model, OpenAI has introduced GPT-5.4 Thinking in ChatGPT. It is designed for more complex prompts and provides a visible overview of the model’s reasoning process as it works through a problem. Users can adjust instructions mid-response, allowing the AI to guide the AI to a more desirable outcome without having to restart the conversation.
GPT-5.4 is also designed to handle longer and more complex tasks, storing information across multiple steps and advanced workflows. These improvements could be particularly useful in coding tools like OpenAI Codex, where the model can help automate large or time-consuming development tasks.
The model is now rolling out to ChatGPT users on web and Android, with iOS support expected soon. OpenAI also offers a GPT 5.4 Pro version for enterprise and academic customers who require maximum performance for complex workloads.
