Why GPT-5 Is brilliant—
and why Its launch sparked backlash

OpenAI’s newly released GPT-5 arrives with bold claims: the smartest, fastest, and most useful model it has ever built. It excels in coding, writing, health advice, and multimodal reasoning, while slashing hallucinations and offering an enormous 400,000-token context window.

In an unprecedented move, OpenAI made GPT-5 the default for everyone, from free users to Plus and Pro subscribers. Premium tiers also get access to GPT-5 Pro, a beefed-up variant with higher limits. For casual users, it’s a straightforward upgrade. For power users, it’s been anything but smooth.

One model to rule them all — sort of

GPT-5 is the company’s first attempt at what it calls a “unified” system, meaning not a single model, but several working behind the scenes, with a “smart router” deciding which to use. If you need speed, it serves a faster, lighter model; if you need complex reasoning, it switches to a heavier “thinking” model. The router learns from your behavior, aiming to pick the right tool automatically.

For everyday users, that’s a welcome simplification. Gone are the confusing lists of model names. But in this simplicity lay OpenAI’s first major misstep.

The backlash over model choice

Before GPT-5, users could choose their preferred model: GPT-4 for conversational work, 3 for deep reasoning, and so on. GPT-5 removed that choice, letting the router decide. Power users quickly revolted.

The reaction was swift and emotional. Some users weren’t just frustrated; they were grieving. Over time, people have developed strong attachments to specific models like GPT-4, using them as trusted collaborators, creative partners, and even emotional companions. When OpenAI deprecated those models overnight, it sparked a kind of digital bereavement.

CEO Sam Altman admitted the rollout was mishandled and promised partial reversals, acknowledging the unexpected depth of users’ emotional bonds with the technology.

Rate limits and the...