Not quite the upgrade people wanted<\/b>
\nOpenAI\u2019s debut of GPT-5 came with many promises of a smarter, more effective AI platform capable of fulfilling many of the dreams of AI enthusiasts. The company was so confident in GPT-5 that it became the default model for ChatGPT, while the previous flagship GPT-4o model was removed from the AI chatbot.
The result went over poorly, judging from the thousands of complaints registered on Reddit and other social media platforms. From power users who relied on GPT-4o to people who thought of the vanished AI model as a friend or more, the blowback was fierce. In fact, it was loud enough that OpenAI reversed course and said GPT-4o will return as a selectable option for Plus subscribers.
The company still says the direction is the same, with plans to unify ChatGPT into a simpler, auto-switching experience. Still, the new model’s issues and the way people responded to the company plucking GPT-4o from the menu of models show that not everyone subscribes to OpenAI’s vision of AI’s future. As for what led to so much outrage, the answers are varied, but all are tied together along similar themes. None of the complaints limited themselves to just one particular irritant, but they all pointed toward the same issues.Clipped clips
One of the top-voted r\/ChatGPT posts summed up the frustration over the new model as less impressive than its predecessor. \u201cShort replies that are insufficient, more obnoxious AI stylized talking, less \u2018personality\u2019 and way less prompts allowed with plus users hitting limits in an hour\u2026 and we don\u2019t have the option to just use other models.\u201d
This isn\u2019t just about length; it\u2019s about the perceived ambition. The style complaint is telling, too, since it implies the safety and style filters are turned up, and the model\u2019s personality has been flattened into corporate vanilla.
The prompt-limit complaint is essentially about trust as well \u2013 if Plus users can hit their cap that quickly, they likely feel like they\u2019ve paid for something that\u2019s already rationed. Underneath all that is a control issue: people want to decide how much of the AI\u2019s brainpower and chattiness they\u2019re renting for a session.Rigid thinking
Another major complaint had to do with what many users said was the rigidity of how GPT-5 followed its algorithm. Many said they found GPT-5 to be less creative as a partner than GPT-4o. One thread started with someone explaining their occasionally chaotic brainstorming approach and how it worked with ChatGPT’s different models.
\u00ab\u00a04o could keep up with me perfectly. It would go deep on A, then go deep on B, and then put them together in a way that made sense. GPT-5 feels like it gets stuck on A and can\u2019t follow me to B and back smoothly. Its thinking is more linear and rigid\u00a0\u00bb, the Reddit poster explained.
\u00ab\u00a0That\u2019s fine for math or engineering because the reasoning is solid. But for brainstorming or organizing messy ideas, it just doesn\u2019t work as well. It\u2019s lost the ability to hold multiple threads and connect them naturally.\u00a0\u00bb
The rigidness also points to a deeper design choice by OpenAI for making GPT-5 better at being correct by limiting how much of a tangent it might go on. That might make sense for code, math, or legal drafting, but it alienates the people who come to ChatGPT for wild ideas and sprawling narratives.Bland AI
For years, OpenAI has insisted ChatGPT isn\u2019t a person. That doesn’t mean people didn’t anthropomorphize the AI, though, for better or worse. Yet, you don’t have to want to believe ChatGPT is sentient to want to enjoy interacting with it. As part of a larger list of issues, one Reddit user explained that they find GPT-5 \u00ab\u00a0creatively and emotionally flat\u00a0\u00bb and \u00ab\u00a0genuinely unpleasant to talk to.\u00a0\u00bb
\u00ab\u00a0Where GPT-4o could nudge me toward a more vibrant, emotionally resonant version of my own literary voice, GPT-5 sounds like a lobotomized drone. It\u2019s like it\u2019s afraid of being interesting\u00a0\u00bb, they wrote. \u00ab\u00a0GPT-5 just sounds tired. Like it’s being forced to hold a conversation at gunpoint. I\u2019m not asking it to be my friend: I just don\u2019t want it to feel like I\u2019m dragging it through every prompt.\u00a0\u00bb
What these complaints really mean is that GPT-5 doesn\u2019t mirror back the user\u2019s tone the way 4o did. Personality isn\u2019t just about \u201cfun\u201d \u2013 it\u2019s about rapport. Lose that, and even technically correct answers feel transactional. For creative professionals, rapport isn\u2019t fluff \u2013 it\u2019s what makes the AI feel like a partner instead of a tool.Model choice
This was the complaint that got OpenAI to blink. One Reddit poster pointed out the hallucinations and forgetfulness of GPT-5 as the start of the problems.
\u00ab\u00a0It feels unstable and inconsistent. GPT-4o was sharp, focused, and reliable\u00a0\u00bb, they wrote. \u00ab\u00a0I did not subscribe to be part of a forced A\/B test with no way out. I want GPT-4o back. I want the option to choose.\u201d
As another put it, \u201cTo abruptly remove the old models and model picker, not even giving us a choice, was a bad move.\u201d
In the end, none of these complaints exists in a vacuum. They\u2019re not just about length limits or creativity; they\u2019re about trust and control. The removal of choice was the most annoying issue to many because it turned what could have been framed as an optional upgrade into a compulsory migration. That\u2019s why the loudest complaints married issues with GPT-5 with a demand for GPT-4o.<\/p>\n