Hi,
Has anyone recently noticed a rapid decrease in ChatGPT’s performance (all models)?
About a month ago, I was super happy with ChatGPT. It really helped me greatly with coding and other work-related tasks. However, over the past three weeks, it has become so bad that I can barely use it anymore. I even feel that the GPT-3 model was better.
Does anyone know more about this?
I’m seriously considering canceling my subscription and testing other LLMs.
Best,
Bronko
I have been a fan of ChatGPT-4 for any in-depth discussions, using gpt-4o & mini only for background services (RAG, parsing, transforming) and ChatGPT4-o for quick drop-in chats.
I was surprised to see that ChatGPT-4 is now listed as a “Legacy Model”.
There has been a lot of optimizations leading to much less cost & I believe the focus is on casual, conversational, actionable models (Think Google Home & Alexa on steroids).
Apparently the “voice update” is coming soon, which I was reading has sometimes a latency response time of <400ms I believe immediately after finishing speaking. Absolutely insane. I have dabbled myself a bit in these AI conversational voice assistants and the best I could do was a second and it was definitely not production-ready.
Going back on track, I’m hoping (and somewhat seeing) ChatGPT-5 around the corner which apparently has PhD-level intelligence and hopefully will also mean that we can return to having powerful, in-depth conversations. So these massive coding projects will be a bit more steerable and coherent.
tl;dr The focus seems to be on casual conversational models & I think there is a general consensus that gpt-4o & mini just don’t quite hit the mark for in-depth coherent technical conversations, but GPT-5 will hopefully be released soon.
Understood but I really got way worse. Lots of logic and context mistakes. I use it for coding everyday but it is giving so many wrong answers that I’m now quicker again to code everything by myself…