GPT4-Turbo is Worse than GPT4
Straw poll for anyone who uses this stuff: the new GPT4 is way dumber than the old GPT4, right? Just curious what others think. For coding at least, it’s way, way better. The new knowledge cutoff means you don’t need RAG, (and there may be more human feedback in the turbo model), threads and the price reduction mean a big reduction in cost (I was averaging about $400/mo in spend, now it’s about $140 [est. based on the daily limit]), and the speed means iterating with the AI is realtime instead of asyc (I used to tab out while waiting for completions). For coding it’s a game changer. For context I’m using Robocoder.app which eats a lot of tokens. Can you give some qualitative examples?