ChatGPT is getting another upgrade, and this time it’s going to GPT 5.4, just days after the release of GPT 5.3 Instant. OpenAI says this new release brings together its “recent advances in reasoning, coding, and agentic workflows into a single frontier model.”
Some of the biggest updates to GPT 5.4 are coming to professional tools, such as improvements to AI-generated spreadsheets, documents and presentations, but it also sees improvements to search and how you can interact with chatbots.
A major change is OpenAI’s next step toward fully autonomous, agentic technologies: the model can use local computer resources, enabling tools to complete complex tasks in applications.
GPT 5.4 can write code to run the computer, which can respond to mouse and keyboard commands depending on what is captured in the screenshot. This means developers can take better advantage of GPT 5.4 to build agents that run other services with limited human interaction.
A benchmark called OSWorld-Verified, designed to monitor AI’s ability to navigate desktop environments, found that GPT 5.4 scored 75%, up from 47.3% with its GPT 5.2 model. It also beats the average human result on the same criterion of 72.4%.
Inside OpenAI’s chatbot, you’ll find Model GPT 5.4 Thinking, which now lets you adjust mid-response while generating a response. If it gets your question wrong or you change your mind about the direction of a query, you can now intervene to align it more closely with the answer without starting over. The feature is available immediately on Android and ChatGPT’s website, and is coming soon to the iPhone app.
Get our best stories!
Your daily dose of our top tech news
By clicking Sign Up, you confirm that you are 16+ and that you agree to our Terms of Use and Privacy Policy.
Thanks for signing up!
Your membership has been confirmed. Keep an eye on your inbox!
ChatGPT 5.4 also allows deep web research, which OpenAI says is particularly helpful for “highly specific queries, better preserving context”, especially when answering long queries.
For professional tasks, OpenAI says it has improved its ability to create and edit documents, making the files it produces easier to read. Its internal tests showed that spreadsheets designed to simulate a junior investment banking analyst achieved an average success rate of 87.3% with human raters.
A rendering of ChatGPT 5.4 (left) next to an example of GPT 5.2 (Credit: OpenAI)
OpenAI also says that there will be fewer errors and illusions with the new model, calling GPT 5.4 its most realistic model yet. It says, “On a set of deidentified cues where users flag factual errors, GPT-5.4 is 33% less likely to make individual claims false and 18% less likely to make any errors in its complete responses than GPT-5.2.”
Recommended by our editors
GPT 5.4 Thinking Upgrades are now available for Plus, Pro, and Team subscribers, replacing the GPT 5.2 Thinking model. The current version will not be retired immediately, but will be moved to Legacy Models and then removed on June 5th. There is also a GPT 5.4 Pro option in the brand’s API for those with Pro and Enterprise plans. As of now, there’s no word on whether free users will be able to use GPT 5.4 Thinking.
Separately, OpenAI has also introduced a dedicated tool, ChatGPT for Excel, to make it easy to plug your spreadsheets into its models. OpenAI says it will help you use Excel data from workbooks to “run scenarios, and generate output based on cells and formulas.”
Disclosure: PCMag’s parent company Ziff Davis filed a lawsuit against OpenAI in April 2025, alleging that it infringed Ziff Davis’ copyrights in training and running its AI systems.
About our expert
Experience
I have been a journalist for over a decade, having started in tech reporting in 2013. I joined PCMag in 2025, where I cover the latest developments in tech, writing about the gadgets and services you use every day. Be sure to send me any suggestions you think would be of interest to PCMag.
Read full bio
