OpenAI’s GPT-5: brilliant on paper, flat in practice 🤖💥

Cliquez ici pour lire en français

Launched just a few weeks ago with sky-high expectations, GPT-5 was supposed to be the next great leap in artificial intelligence. It came wrapped in promises of explosive creativity, sharper reliability, and unmatched power—a generational shift in how we interact with machines. Instead, what users got feels more like a step back. Disappointment has spread quickly, and OpenAI’s much-hyped model is increasingly being labeled a flop. So, what went wrong?

Performance takes a hit, personality goes missing ❄️

One of the biggest complaints is that GPT-5 feels colder and less engaging than its predecessors. Instead of the warm, imaginative assistant people were hoping for, many describe the model as flat, terse, and emotionally distant. The spark of creativity that defined earlier versions seems to have dimmed, replaced by shorter answers with less nuance and personality.

And that’s not all—users are also pointing out odd mistakes: basic spelling errors, geographic inaccuracies, and clumsy phrasing that you’d expect from a far less ambitious system. For a model pitched as a leap forward in natural interaction, this shift feels like a letdown.

Frustrating limits and messy rollout 📉

Another pain point lies in how OpenAI has managed the rollout. GPT-5’s “Thinking” mode—meant to deliver deeper reasoning—is capped at 200 messages per week. For paying subscribers, that hard ceiling feels like a regression compared to the openness of earlier versions.

Adding fuel to the fire, OpenAI abruptly retired older models like GPT-4o, which many considered more reliable. The backlash was swift enough that the company reintroduced GPT-4o—though now behind a paywall—in what looked like damage control. The entire rollout has left users questioning not just the product’s quality, but OpenAI’s communication and strategy.

What’s next for GPT-5? A maturity check 🛠️

The stumble highlights a larger shift in the AI race. Raw power alone may no longer be enough—how these systems feel to use, and the relationships people build with them, are becoming just as important.

OpenAI, caught between sky-high expectations and the brutal pace of innovation, has found itself in a familiar position: reminding the world that AGI—the mythical “general” intelligence—remains out of reach. As CEO Sam Altman himself put it, GPT-5 is “generally intelligent” but still far from truly general.

The real question is whether GPT-5 can bounce back—regaining trust, refining its personality, and proving that performance and human-like experience don’t have to be trade-offs.

GPT-5: a necessary stumble before the rebound ? 🧐

GPT-5 clearly hasn’t lived up to its billing, but sometimes failure teaches more than success. This turbulence may be less about raw capability and more about an AI learning how to listen, adapt, and build genuine rapport with its users.

Perhaps the next breakthrough won’t come from another leap in sheer power, but from creating technology that actually feels human to interact with.

So, what about you? Have you tried GPT-5 yet? Do you think this disappointment is temporary—or are we watching a turning point in AI unfold?