It's ridiculous to gauge a model's capability solely by whether it can produce a slick-looking frontend. GPT-5 just fixed a canvas bug that Claude and Gemini couldn't solve in two days.
338