OpenAI is quietly tinkering with a small change that could save power users a lot of thumb-tapping: a swipe gesture in the ChatGPT mobile app that lets you switch AI models almost instantly.
If you bounce between GPT-4, GPT-4o, and the free-tier options depending on what you’re doing—quick question, heavy lift, image stuff, whatever—this is the kind of “why wasn’t it always like this?” tweak that actually matters.
A simple swipe that cuts the menu nonsense
The idea is dead simple: swipe horizontally and you get model selection, without digging through the usual navigation. It borrows a move people already know from messaging apps and social feeds—thumb goes left or right, the app obeys.
Design-wise, it’s OpenAI choosing speed over the old-school “everything lives in a menu” philosophy. And for anyone who treats ChatGPT like a multi-tool instead of a novelty, that’s the right call.
Because OpenAI’s lineup is a choose-your-own-adventure now. GPT-4 is typically better for gnarlier, more complex tasks—but it’s heavier. Lighter models handle basic requests faster and cheaper. The whole point of offering multiple models is that they’re different. The whole problem is that switching between them can be annoying enough that people just… don’t.
Why OpenAI is sweating the tiny stuff
This isn’t just a cute UI flourish. It’s a competitive move in a market where the big players are starting to feel interchangeable to regular humans.
Anthropic’s Claude, Google’s Gemini, and the rest aren’t only fighting on raw model performance—they’re sanding down friction everywhere they can. When the tech is close enough, the winner is often the product that feels faster, cleaner, less fussy.
Shave a few seconds off a common action, multiply it by millions of daily users, and you’ve bought yourself something companies love: habit.
The catch: making users choose models exposes the business math
Here’s the tension hiding under the “slick new gesture” headline: the easier OpenAI makes it to hop between models, the more it’s admitting that users have to play traffic cop—deciding which brain to use for which job.
That’s the opposite of the “black box” dream where one model magically handles everything and you never think about it.
And there’s a money angle. If paying users can swipe between premium-grade models and free options with zero hassle, the value proposition gets clearer—sometimes uncomfortably so. When it’s effortless to compare “good enough for free” versus “better for money,” conversions can get harder, not easier. OpenAI has to balance convenience with the economics that keep the lights on.
A small test with big implications for how people use ChatGPT
Right now, this swipe feature is described as being in testing. But the real point of the experiment isn’t whether people like swiping. Of course they do.
The point is what it does to behavior: Do users start switching models constantly? Do they settle into a pattern that favors cheaper models? Do they discover they don’t need the premium one as often as they thought?
A tiny gesture can change the whole rhythm of an app. OpenAI knows that. That’s why this “minor” update is anything but.
