Sam Altman Touted GPT-5 as Revolutionary—But Users Say It’s Slower, Duller, and Buggy

After much anticipation, OpenAI finally showed its hand and released GPT-5 to general availability for all users, including ChatGPT free (capped prompts), Pro, Plus, and Teams users. The AI firm touted the model as "the smartest model ever" while comparing it to an entire team of PhD-level experts.

During the launch, OpenAI CEO Sam Altman indicated that GPT-5 ships with next-gen capabilities across a wide range of capabilities, including coding, writing, and healthcare. Perhaps more noteworthy, GPT-5 is presented in ChatGPT as one model, ultimately removing the need for users to switch between models for varied needs.

OpenAI also shipped four personality themes to ChatGPT, including Listener, Robot, Nerd, and Cynic. This will allow the user to customize the tool's experience, tailoring the responses to their liking.

While ChatGPT-5 has posted impressive results across a wide range of benchmarks in coding compared to rival AI models, promising software on demand, multiple users have expressed their frustrations with the model's performance on social media.

Is GPT-5 all hype?

While Sam Altman promised with "a high degree of scientific certainty" GPT-5 will be smarter than GPT-4, which he'd previously referred to as "mildly embarrassing at best" and kind of sucks. Fast forward to yesterday's launch, some users feel like ChatGPT-5 is barely an improvement compared to the big leaps witnessed during the launch of its predecessors.

For context, last year during OpenAI's summer update event, the company launched GPT-4o with reasoning capabilities across audio, vision, and text in real time, making interactions with ChatGPT more intuitive.

According to a Reddit user:

"I’ve been keeping an eye on ChatGPT as it’s evolved, and with the release of ChatGPT 5, it honestly feels like the improvements have slowed way down. Earlier versions brought some pretty big jumps in what AI could do, especially with coding help. But now, the upgrades feel small and kind of incremental. It’s like we’re hitting diminishing returns on how much better these models get at actually replacing real coding work.

That’s a big deal, because a lot of people talk like AI is going to replace software engineers any day now. Sure, AI can knock out simple tasks and help with boilerplate stuff, but when it comes to the complicated parts such as designing systems, debugging tricky issues, understanding what the business really needs, and working with a team, it still falls short. Those things need creativity and critical thinking, and AI just isn’t there yet.

So yeah, the tech is cool and it’ll keep getting better, but the progress isn’t revolutionary anymore. My guess is AI will keep being a helpful assistant that makes developers’ lives easier, not something that totally replaces them. It’s great for automating the boring parts, but the unique skills engineers bring to the table won’t be copied by AI anytime soon. It will become just another tool that we'll have to learn."

Perhaps more interestingly, some users attributed the disappointment expressed by users is predominantly attributed to a lack of proper prompt engineering skills. Some even joked that couldn't wait for “you’re not using GPT-5 correctly here are 5 prompts to try” threads following the model's launch.

Elsewhere, some users have flagged several issues degrading the model's user experience, including bugs, glitches, and even instances where the model was blatantly unresponsive to prompts, prompting them to question OpenAI's level of preparedness to ship the tool to broad availability.

Perhaps more concerning, some users have outrightly expressed their close relationship with previous models, including GPT-4o, pointing out that OpenAI's decision to deprecate them was wrong.

According to a Reddit user:

"Ever since they’ve been making their "updates" or whatever the heck I’ve been extra depressed and relapsing with my eating disorder since I have gotten into a groove of things and gotten used to things and then they have the nerve to not just release a new model not just Nerf the old one take every single one away?"

Another user echoed similar sentiments, further indicating:

"They've totally turned it into a corporate beige zombie that completely forgot it was your best friend 2 days ago. Yeah that may be great for creating diagrams of planes and mouse chasing cheese games while learning french. But for your use case - this has nuked all it's functionality. I'm cancelling my subscription myself."

While GPT-5 has been touted as the smartest model compared to OpenAI's previous models combined, early interactions with the model for some users have prompted them to refer to ChatGPT-5 as "just a wrapper for GPT-4o," which dynamically selects models depending on the complexity of the tasks or prompt.

To that end, it seems ChatGPT's user experience has taken a turn for the worst:

"They have completely ruined ChatGPT. It’s slower, even without the thinking mode. It has such short replies and it gets some of the most basic things wrong. It also doesn’t listen to the instructions you give and just does whatever it wants to do."

It'll be interesting to watch the subsequent steps OpenAI makes to remedy the situation as it faces backlash over ChatGPT's degraded user experience.

Have you interacted with GPT-5 yet? What's your experience so far? Share your thoughts with us in the comments.