comparison of o1 vs sonnet for converting a Figma design to code is the extra time and cost worth it?

conversion done with @builderio using claude-3-5-sonnet-20241022 and o1-preview-2024-09-12 I hand checked the same prompts on chatgpt and didn't see any major differences in performance or results on the latest o1 there deeper dive on this:

If the code is even a little better then yes, even one small bug will take me way more time to fix

I want to comment on something about it, but I'm still thinking...

The only thing that matters is accuracy.

Also show the end results? If o1’s result is way better, then I would rather wait a minute more

O1 is about 10 times faster than the speed of o1-preview, and significantly better quality, even without “pro mode”, but until we get api access this video will be valid. So meh L take

You're super bearish. Do you always get pixel perfect stuff with your tool? Does the model "see" the visual output it generates? Design->Code is all about iteration for human. You code, you look (!), you correct. When the model can get feedback from rendered code (screenshot) o1 is gonna be a game changer.

o1-preview != o1-pro

Which one has a better end result tho

Very strange comparison :). I like this "Accuracy ? : ?". For simple frontend tasks of course it's enough to use faster model. You could try the same with chat-4o model. o1 not about speed, it's about accuracy and reasoning.