The launch of GPT-5, announced as a major step toward artificial general intelligence (AGI) by OpenAI and its CEO Sam Altman, was supposed to revolutionize the AI landscape.
After months of teasing, expectations were sky-high. Yet, far from sparking enthusiasm, GPT-5 has disappointed a large part of the community, which much prefers its predecessor, GPT-4o. A look back at a chaotic launch and the reasons for a massive rejection.
A cold reception for GPT-5
As soon as it was deployed, GPT-5 drew sharp criticism. Contrary to what OpenAI’s communication suggested, the model failed to convince. Many users describe it as “austere,” “cold,” and “clinical,” lacking the warmth and conversational fluidity that made previous versions so charming, especially GPT-4o.
The latter, praised for its versatility and natural tone, seems to have set a standard that is difficult to surpass.
The problem is worsened by a controversial decision: OpenAI removed access to GPT-4o for ChatGPT users, forcing the adoption of GPT-5. This approach, which imposes a single model supposed to adapt automatically to each request, has caused an outcry. Users miss the ability to choose their model, a feature that offered them more control and flexibility.
For many, this unification is seen as a regression, stripping the AI of the personalization that was its strength.
Read also on this topic: Which ChatGPT to choose? Between models 3.5, GPT4, 4o, o1, o3?
ChatGPT 4o vs GPT-5: A revealing technical showdown!
GPT-5, touted by OpenAI as a leap toward AGI, disappoints with its technical performance compared to ChatGPT 4o.
Where GPT-4o shines with its optimization for complex tasks like natural text generation, contextual query processing, and reduced latency, GPT-5 suffers from a rigid unified architecture that prioritizes automatic model selection at the expense of personalization.
Users report less accurate responses from GPT-5, particularly in creative and analytical tasks, with sometimes longer response times. In contrast, GPT-4o maintains unmatched robustness thanks to its fine-tuning for versatility and its more intuitive interface. This technical disparity highlights a major pitfall: GPT-5’s innovation seems to sacrifice practical efficiency for a theoretical vision!
GPT-4o, the undisputed favorite?
User feedback is unanimous: older models, notably GPT-4o, but also GPT-4.1 and GPT-4.5, remain more effective for tasks like creative writing, content generation, or even informal conversations.
Their tone, judged to be friendlier and more engaging, contrasts with GPT-5’s more mechanical approach. This preference isn’t just a matter of style: GPT-4o’s performance, especially in complex scenarios, is often considered superior.
Fortunately, alternatives exist! For example, our Yiaho platform continues to offer GPT-4o, allowing users to bypass the restrictions imposed by OpenAI. This availability has actually amplified the frustration, as ChatGPT subscribers feel cheated by the abandonment of their favorite model.
See also: ChatGPT 5 vs Grok 4: Which is the best AI?
OpenAI’s response: A timid mea culpa
Faced with the wave of discontent, Sam Altman spoke out during a Q&A session on Reddit. Acknowledging that the GPT-5 rollout was “more chaotic than expected,” he hinted that access to GPT-4o could be restored for ChatGPT Plus subscribers. This announcement, while welcome, doesn’t erase the feeling of a poorly thought-out strategy.
Forcing the adoption of a single model without offering an alternative clearly underestimated users’ attachment to the flexibility and quality of previous versions.
A lesson for OpenAI?
This failed launch raises crucial questions about how AI companies manage expectations and transitions between models. By trying to impose GPT-5 as an all-in-one solution, OpenAI seems to have neglected a fundamental principle: user experience takes precedence over technological ambitions.
User feedback shows that innovation is not measured only in terms of raw power, but also in terms of user-friendliness and adaptability. For now, GPT-4o remains the users’ darling, and pressure on OpenAI to restore access to this model is only growing.
In the meantime, our Yiaho platform fills the gap, offering a valuable alternative for those who refuse to compromise. The GPT-5 fiasco serves as a reminder of an essential truth: in the race for AGI, listening to users is just as important as pushing technological boundaries!
Source: LesNumériques
