OpenAI Brings GPT 5.4 AI Model With Agentic Powers That Work On Computer

Last Updated:March 06, 2026, 16:30 IST

OpenAI is adding more AI agent work with the new GPT version that will be offered to the premium and Plus subscribers.

OpenAI has been busy in the last few days and now we are moving to the GPT 5.4 AI model which brings more agentic powers that can be summoned on computers. There are different versions of GPT 5.4 including Thinking and Pro which ChatGPT Plus, Team and Pro subscribers can access while the latter is available for the Enterprise and Pro premium users of the AI chatbot.

OpenAI claims both the new versions have been improved to deliver better coding, reasoning and more importantly agentic processes. But the main highlight of GPT 5.4 is its ability to handle AI agents in computer-use. 

More AI Agents With GPT 5.4 

What does OpenAI mean by computer-use for AI agents? The company is letting the GPT 5.4 AI model operate a virtual computer and all the tasks that you use the systems for. The new version is also capable of writing the code that will make the computer handle tasks, control it with a mouse and understand the keyboard actions. 

OpenAI claims the new GPT model got a score of 75 percent on the OSWorld-Verified benchmark test which shows early promise of the AI model and its understanding of commands in the desktop environment. AI agents are the real focus of most AI companies, including OpenAI, and this model is heavily built to handle the power-centric demands, wherein you can order the AI agent to browse the web, make orders online, and other tasks. 

But having AI agents handle more important tasks means the security nature of the GPT 5.4 will also be heavily discussed. OpenAI says the new model has been classified to handle advanced cyber concerns, monitors any possible threats to the systems and handles the controlling of the machine in a secure manner. 

OpenAI is also facing the risk of alienating millions of users. The ChatGPT uninstalls have gathered pace this week, especially after OpenAI confirmed its plans to work with the Department of Defense (DoD) of the US government.

This is a big deal for Anthropic after all the recent criticisms publicly made by its chief Dario Amodei, who has refused to fully comply with the government terms and access of AI data for various purposes. Industry reports confirm the change in trends as people switch from ChatGPT to Claude and the numbers are staggering to say the least.

News tech OpenAI Brings GPT 5.4 AI Model With Agentic Powers That Work On Computer

Read More