Less than 24 hours after OpenAI’s much-anticipated GPT-5 launch, the internet is abuzz with one key message: Users Are Not Happy with ChatGPT 5 and Call It Horrible. The new model, hyped for weeks by CEO Sam Altman as a “PhD-level” upgrade, is facing a storm of criticism from paying and free users alike.
Launch and Initial Reactions
During a one-hour livestream, Altman and the OpenAI team unveiled GPT-5’s supposed breakthroughs — sharper reasoning, multi-step processing, and enhanced personalization features. The event was positioned as a milestone in AI evolution, building on the success of the GPT-4o series.
However, excitement quickly soured. Within hours of the announcement, Reddit threads began filling with complaints. One titled “GPT-5 is horrible” shot to nearly 3,000 upvotes and over 1,200 comments in less than a day. Many participants didn’t just criticize GPT-5’s performance — they openly demanded the reinstatement of GPT-4, which was quietly removed from the model selection menu for most users.
“Shorter replies, slower thinking, less creativity — this feels like a downgrade, not an upgrade,” one Redditor wrote.
Common Complaints: From Speed to Substance
Early adopters have flagged several recurring problems:
- Poorer response quality – Many claim GPT-5’s answers are less detailed and lack the “human-like personality” that made earlier models engaging.
- Slower output – Some report noticeable lag, especially when using the new “Thinking” mode for complex reasoning tasks.
- Feature limitations for paying subscribers – GPT-5’s “Thinking” mode is capped at 200 messages per week for ChatGPT Plus users, frustrating those who rely on the tool heavily.
- Loss of model choice – GPT-4 and GPT-4o are no longer standard options for many users, replaced entirely by GPT-5’s self-selecting reasoning engine.
These limitations sparked accusations of “AI shrinkflation” — the idea that users are getting less capability or choice despite the product being branded as a major upgrade.
The “Shrinkflation” Debate
The term “shrinkflation,” often used in consumer goods, has been borrowed by AI critics to describe GPT-5’s rollout. While benchmark scores suggest the model is more advanced under the hood, many say its real-world usability is worse.
“On paper, it’s smarter. In practice, it’s more annoying,” one early user complained on X (formerly Twitter).
Others were irritated by OpenAI’s decision to remove older, reliable models entirely instead of keeping them available for those who preferred their style and output.
Expectations vs. Reality
The backlash is amplified by the intense marketing campaign leading up to launch. Altman fueled speculation with cryptic hints, including a Star Wars-style teaser that led some to expect a revolution in AI capability.
When GPT-5 finally arrived, the leap felt smaller than promised. While performance metrics are objectively higher — especially in reasoning benchmarks — for many users, the differences are subtle and fail to justify the loss of older features.
This isn’t the first time OpenAI has faced a mismatch between marketing and user perception. But given ChatGPT’s massive user base and loyal following, any perceived downgrade triggers an especially strong reaction.
OpenAI’s Response
Facing mounting pressure, OpenAI has begun damage control. Altman acknowledged some performance issues in the initial rollout, blaming them partly on a “router bug” that mismanaged requests between GPT-5’s different reasoning modes. The company says it is now:
- Investigating performance slowdowns
- Evaluating user feedback for personality tuning
- Considering restoring GPT-4o for Plus subscribers
Early signs suggest the company might walk back some changes, especially around model availability, to calm its frustrated community.
Industry Reactions
Tech analysts are split on the backlash. Some believe it’s simply a case of users needing time to adapt to a new style, while others see it as a cautionary tale about overhyping product launches.
“GPT-5 is better in certain measurable ways, but worse in others that matter to human users,” said one AI researcher. “Performance benchmarks don’t always capture user experience.”
Rival AI providers are already seizing the moment, with some promoting their own models as more “reliable” or “consistent” alternatives. This competitive pressure could push OpenAI to accelerate improvements.
The Bigger Picture: Trust and Consistency
Beyond speed and features, the GPT-5 controversy highlights a deeper issue — trust. Millions of people now use AI tools not just as search engines, but as collaborators, writing partners, and research aides. A sudden shift in tone, output quality, or availability can feel like losing a trusted colleague.
If OpenAI hopes to maintain its leadership position, it may need to rethink how it introduces changes, ensuring they enhance rather than disrupt user workflows.
ALSO READ: Does Instagram Kill Your Views for Using “Link in Bio”?
What Happens Next?
The coming weeks will be crucial. If OpenAI restores GPT-4o, addresses performance issues, and improves GPT-5’s warmth and creativity, the model could yet win back skeptics. If not, the “GPT-5 is horrible” narrative might stick — at least until the company rolls out GPT-5.5 or GPT-6.
For now, the message from the community is loud and clear: they expect upgrades to feel like upgrades. And if they don’t, users won’t hesitate to voice their disappointment.
