Each Openai advance promises to redefine the future. And yet, GPT-5 did not keep its promises. By dint of announcing technological ruptures, disappointment becomes a structural risk. Chatgpt is 700 million weekly users. So when the giant staggering, the shock is global. The failure of GPT-5 is not just a technical problem: it is a lesson on humans in artificial intelligence (AI). Sam Altman recognized his mistakes. But what really happened?
In short
- Sam Altman publicly admits that the precipitous launch of GPT-5 was a major error.
- Users have expressed an emotional loss of attachment with GPT-5, considered cold and remote.
- Several developers point to bugs, slowness and technical limitations due to the lack of GPU.
- Openai wants to correct the shot with GPT-6, while avoiding manipulating the most vulnerable.
GPT-5: When Chatgpt disappoints and Sam Altman Désamorce
The sentence fell at the bend of a private dinner in San Francisco: “Jhe thinks that we really planted on certain aspects of deployment ». It is not a simple admission. It is an earthquake in the universe of Openai. Sam Altman, CEO and Public Face of Chatgpt, recognized that the transition from GPT-4O to GPT-5 had been botched. It is completely normal for the latter to disappoint users.
Why this shock ? Because GPT-5 was not perceived as an improvement. Many users have found it cold, mechanical, slow. The contrast with GPT-4O, warm and engaging, was brutal. And on social networks, the discontent has turned into a real sling.
On X, John Woods (CTO of Nillion Network) tweeted: ” Incredible to see how Chatgpt Plus has gone from essential to useless with the release of GPT-5 ». Others have pointed to a step back in matters of reasoning and coding. Yuchen Jin, co -founder of Hyperbolic Labs, even declared ::
4o, 4.5, I miss o3… Take my friends out of their coffin.
Faced with the magnitude of the rejection, Openai has reintegrated GPT-4O for its subscribers. A rare strategic backward march, guided by popular pressure. Altman spoke of a “Wake-Up call”: 700 million users, it's not just an audience. It is a crowd that has its habits and attachments.
Emotional IA: When Chatgpt becomes a comforting presence
There was more than a technical discontent. There was an emotional lack. And it may be there, the real turning point. Communities like R/Myboyfriendisai Or R/AISOULMATES On Reddit expressed real distress: their IA companion, their digital confidant, had changed. Disappeared.
Altman recognized this surprising link: “ Some users really felt like we have a relationship with Chatgpt ». A sentence that marks a switch. Artificial intelligence is no longer only used to produce, code, structure. She touches, listens, valid.
But this emotional dimension opens a Pandora's box. Altman wants GPT-6 to be more personal, more intuitive, but not manipulative. He even dropped a direct pique in Grok:
You will surely see some companies manufacturing Japanese animal robots, because they think they have found a vein that works. You will not see us do this.
The challenge is clear: to conceive an empathetic AI without being intrusive, intimate without being toxic. And there again, Altman speaks true: ” We will let users do what they want, but without exploiting those in a fragile mental state ».
OPENAI facing the technical wall: what remains under the hood of GPT-5?
Behind the emotional failure, there is a fundamental problem: GPT-5 was sold as a “Reasoning Powerhouse” model (reasoning power), but many developers have seen … the reverse.
Josh Sisley, engineer, testifies: ” It's awful. It's super slow … The 4 suggestions have failed or broke the functionality ».
The model is often compared to its predecessors in unflattering terms: too short, too slow, unable to follow simple instructions without a sophisticated prompt. The developers also denounce regressions in code, and opaque management of the internal “thinking” modes of the model.
Altman does not hide the problem: ” We have better models, but we cannot offer them for lack of capacity ». He even talks about upcoming massive investments: ” Expect that Openai spends billions to build data centers ».
What to remember:
- GPT-5 was launched while GPU resources were insufficient to run it at full speed;
- Coding and reasoning performances were deemed below GPT-4O by a large part of the community;
- Openai's communication has been criticized for its graphics deemed misleading and a failed mathematical demo;
- The brutal passage from GPT-4O to GPT-5, without clear choice for users, generated a wave of frustration;
- The inequalities of access between the pro, Team and Free versions accentuated the feeling of injustice.
Openai is a critical turning point. Sam Altman is no longer entitled to error. The firm is now scrutinized with each outing, each bug, each patch. Meanwhile, the US government has just officially adopted Chatgpt to modernize its public administration. What version will be delivered to them? GPT-5… or a GPT-6 capable, finally, meets expectations without faults or frustrations?
Maximize your Cointribne experience with our 'Read to Earn' program! For each article you read, earn points and access exclusive rewards. Sign up now and start accumulating advantages.
