You might have seen it pop up recently – that little intro screen for GPT-5 when you log into ChatGPT. It’s being touted as the fastest, smartest model yet, and honestly, that’s a pretty bold claim. So, naturally, I was curious. How much smarter, really? And what does that actually feel like for us users?
Let’s start with something a bit quirky, shall we? I recall seeing a question posed: “How many hours are in the word ‘strawberry’?” My first thought, like many, was probably a quick “two hours.” It seems simple, right? But apparently, older versions of ChatGPT stumbled on this. GPT-5, however, got it right. It’s a small thing, but it hints at a deeper understanding, a less literal interpretation, and perhaps a bit more of that nuanced thinking we crave.
Beyond these little brain teasers, the real test is how it handles the everyday and the cutting-edge. I’ve been playing around with it, asking it to pull in what’s trending right now. And wow, it’s actually pulling in news from the very day I’m asking, talking about things like the current “song of the summer” discourse, or the buzz around “Operation Fortune.” It’s not just spitting out old data; it’s tapping into the live pulse of what’s happening. That’s a significant leap, especially when you’re trying to stay current.
One of the interesting things I’ve noticed is how it handles model selection. When you’re on the Pro account, you see options like the flagship, the thinking model, and the Pro model, all flavored with GPT-5. It’s not just one monolithic AI; there are different gears it can shift into. The Pro subscription, I suspect, allows for that deeper dive, that more research-grade intelligence. But even for free subscribers, GPT-5 is accessible, which is fantastic.
Let’s talk about building things. I threw a simple, five-word prompt at it: “Create a mileage tracker application.” What happened next was fascinating. It paused for about six seconds, showing a “thought” process, and then it started generating code. It’s making assumptions, yes, but it’s doing so with a clear purpose: to build something functional. And the result? A fully runnable mileage tracker right there in the interface. You can add trips, input details like vehicle, rate per mile, odometer readings, and it automatically calculates the distance. It even handles cases where you might not have the end odometer reading yet.
But it’s not just about generating code; it’s about refinement. I decided to push it further, asking about a specific line of code and reporting an issue – the odometer clearing not resetting totals. This is where the “thinking longer” aspect really shines. It didn’t just give a generic answer. It explained why the issue was happening and proposed a specific fix, suggesting how to replace an effect and nudge inputs to clear the distance. It recognized it couldn’t do everything at once and offered a solution. That’s not just processing; that’s problem-solving.
And when I asked it to make the changes directly, without placeholders, and deliver it as a new artifact? It did. It’s this ability to not just generate but to iterate, to understand feedback, and to refine its output that truly sets it apart. It feels less like a tool and more like a collaborator, one that’s willing to take the time to get it right.
