Nuked 
Hey everyone, Nuked here! Today, we’re diving into the colorful world of ChatGPT and its ever-complicated model picker. Buckle up for some tech fun!
When OpenAI launched GPT-5 last week, the goal was clear: make the ChatGPT experience smoother by introducing a “one size fits all” AI model that would handle everything seamlessly. This new model was supposed to act as a smart router, choosing the best AI to respond to your questions without you having to lift a finger.
However, things didn’t go exactly as planned. OpenAI’s CEO, Sam Altman, announced new settings for GPT-5—”Auto,” “Fast,” and “Thinking”—that users can choose from the model picker. The Auto mode is designed to work like the original model router, but users also have the option to bypass it and pick faster or slower responses manually. Most folks will probably stick with Auto, but now there’s a bit more control in your hands.
Along with these new modes, Altman mentioned that paying users can still access older AI models like GPT-4o, GPT-4.1, and o3, which had been deprecated just last week. Interestingly, GPT-4o now appears by default in the model picker, and users can add others through settings. The team is also working on a warmer personality for GPT-5, hoping to make the bot more friendly without losing that crucial balance.
But here’s the twist: despite the promising features, ChatGPT’s model picker still looks rather complicated. The rollout has faced some bumps, especially with GPT-5’s model router not working flawlessly for everyone. This led to frustration among users, some of whom even felt the AI models’ responses were less satisfying than previous versions. Altman has acknowledged these issues and promised quick improvements, emphasizing the challenge of matching AI models to user preferences.
It turns out many users are quite attached to certain AI models—so much so that some have even held a funeral for Claude 3.5 Sonnet after it was taken offline. The process of aligning AI behavior to individual tastes is still a work in progress, and the terrain remains pretty bumpy.
All in all, GPT-5’s model router hasn’t fully won over everyone, and the model picker retains a touch of chaos. The high hopes for GPT-5 led to some disappointment, but OpenAI remains committed to refining the experience, aiming for more personalized AI interactions in the future.
