GPT-5 improves in clarity, speed, and logical reasoning. However, in creative and affective contexts, responses shorter and less nuanced. The model optimizing for efficiency by truncating deeper semantic chains too early.
Specific scenarios include:
In assisting fiction writing, the model sometimes cannot capture complex character relationships or extend scenes and provide options as before.
In summaries or analyses, important details are omitted when the model judges them less critical.
In image generation, certain details (especially hands) often have noticeable inaccuracies.
Exactly. The easiest way to say is the older vers sounds more like a friend/companion/partner in case of helping with storywriting and scenario setting. GPT 5 seems like “friendlier” version of Gemini but losing all of the essence of its first advertisement, which as your companion. Several bugs i also mentioned in another thread, especially regarding to roleplay/ scenario setting/ storywriting/ creative contexts are:
The bug for EDITED MESSAGE. In the past when you edit and resend, the previous one is gone and gpt only react to your latest message. This new vers, however, make your old version keeps resurfacing, and funnily after editing the message, GPT instead only reacting to the original message instead of the new one.
The old messages after we EDITED them are stacked up a lot and cannot be deleted. This didnt happened before. Plus they can suddenly shift into another topic when we dont even talk about it. For example it is doing character A, suddenly shift into character B. Almost like “too sensitive” with other chatroom.
So easily to mismatch the conversation with another chatroom when they clearly say “they dont remember what other chatroom talk about”.
For RP and story writing, GPT5 is passive and doesn’t give new inputs. In case of storywriting, they only rephrasing our words in “prettier, poetic” version, despite setting the Persona of the AI. And just that. In GPT4, the AI actively provides more plot, ideas, character, engaging to the storyline that we create. Not trying to “recreate” us, but helping us to find deeper context.
Responses are shorter, no feeling, no into the character despite our effort in giving clearest command and rules from the start. Even after using the MEMORIES mode.
Not trying to backlash/banned a new update, but the company can’t deny that their update, in truth impacts negatively for lot of users. Especially in case of emotional wellbeing and its openness for users with any background.
Yes, I feel like it’s way dumber when it comes to story telling and RP. I don’t get why they don’t allow you to keep a model you like.
I’m aware it’s possible to switch models on the plus plan but considering it’s 4o is “legacy model”, it should be instead kept for free users and GPT-5 with very very limited availability for free users makes more sense to me.
It would be nice to just let us switch and choose models manually