Anthropic recently updated its flagship AI model, Opus. Opus 5.5 is more token-efficient and autonomous than its predecessor, Opus 5. Since OpenAI also released its latest and most powerful AI model, GPT-6 Astra, it’s worth testing both models side by side to see which is best for product design tasks.
In this article, I will test GPT-6 Astra and Opus 5.5 on 2 tasks: building a new design from scratch and introducing a change in the generated design.
Test preconditions
For my test, I will generate a prototype of a mobile app for food ordering.
Below is the complete prompt I will use. It outlines the product type, the requirements for functionality and styling, and the level of detail I expect from AI.
Code interactive prototype of a home screen for a modern food ordering mobile app.
The screen should help users quickly discover restaurants, browse food categories, find relevant offers, and reorder familiar meals.
Now this is the kind of AI comparison that makes you stop scrolling! 👀 I liked seeing how differently the two models approach product design—really interesting to think about what that means for designers going forward.
I really appreciate you putting together this comparison as an article. I liked the opus design better than Astra and I'm glad to learn that opus is way token efficient than astra. Let's say...if we continue to implement this design what dev technology opus will use by default to generate the code?
