Have you ever wondered what would happen if we swapped out GPT 4 with GPT 5 as the thinking model for computer use agents? Well, wonder no more! A recent experiment has shown that GPT 5 can pull through with flying colors.
The experiment used the same tasks and grounding model as before, but with GPT 5 as the thinking model. The results were impressive, with GPT 5 navigating to a random URL and playing a game until it reached a score of 5/5. This task was set up by having Claude generate a random app from a predefined list of prompts, including multiple choice trivia, form filling, or color matching.
What’s even more exciting is that you can try it out for yourself! The code is available on GitHub, and the documentation provides a step-by-step guide on how to get started. This is a huge step forward in making AI agents more practical for everyday use.
Imagine being able to tell your computer to perform tasks without having to lift a finger. It’s not just about convenience; it’s about opening up new possibilities for people with disabilities or those who struggle with certain tasks. GPT 5 has the potential to revolutionize the way we interact with our computers, and it’s an exciting time to be a part of this journey.