3.5 is completely neutered it always gives me short answers. Good to know it’s improving in GPT4-Turbo.
Discussion
Deffo. GPT-4 Turbo can give longer answers with 2x less compute power vs GPT-4 according to the press release.
No way to verify that scientifically ofc but it feels about right judging by UX. Even GPT-4 was neutered a bit on mobile. I noticed it was using Turbo when I started getting long answers even on my phone. I legit thought it was still only available via the API until I noticed the difference on mobile and asked it if it was using Turbo now. So I seriously do notice a real difference.