I’ve been with GPT before the public craze began, before people in general knew about Chat GPT.
Since GPT removed some of the older models that were working much better than the new 5-models, the output quality has been so poor, so bad, so unreliable, it’s not worth subscripting to anymore.
I tried different models (Google’s being one of the most unreliable), and finally ventured to Claude. The difference is night and day. While GPT can’t work on doc’s or excel without hours on end of mistakes and poorly configured files, Claude spits them out in seconds, without any errors. GPT can’t output quality responses anymore, can’t remember what it’s asked to remember, and the replies feel like a constant stream of high-school nonsense. GPT’s highly exaggerated increasing restrictions on any subject and content, makes it fairly useless for anything but recipies and general content.
The only downside with Claude is that you get constant “reached your limit” notifications when you use it very little, even after one prompt, and have not even come close to reaching your limit (clearly a development error they ignore to pay attention to since this is the number one complaint with Claude), but the output quality is far beyond GPT.
If GPT won’t improve and hire better engineers I will not be returning soon. And I think many people will leave GPT. The model started to get good with 4.0 but models after that went downhill and are just getting worse, with the 5-models being the worse of all.
GPT at its current versions and format will not survive competitors, in my opinion. Starting out as the best LLM it’s become fairly unusable.