8/10/2025

Then vs. Now: A Side-by-Side Output Comparison of GPT-4o & GPT-5

I've been playing around with AI for a while now, & I gotta say, the pace of change is just WILD. It feels like just yesterday we were all blown away by GPT-4o. It was fast, it was smart, & it felt like a HUGE leap forward. But, as with all things in the world of AI, the next big thing is always right around the corner. & that next big thing is GPT-5.

OpenAI's latest model has been out for a little bit now, & I've had some time to really put it through its paces. So, is it worth the hype? Is it really that much better than GPT-4o? Let's break it down, then vs. now, & see what's what.

The Big Picture: What's Changed?

First off, let's talk about the architecture. GPT-5 isn't just one model; it's a whole family of them. There's a "regular" GPT-5, a "mini" version, & a "nano" version. & for the real heavy lifting, there's a "thinking-pro" model that's designed for super complex tasks. This is a pretty big shift from the one-size-fits-all approach of the past. It's like OpenAI is admitting that not every task needs a sledgehammer.

The way you interact with ChatGPT has also changed. The model picker is gone. Instead, there's a smart router that decides which model to use based on your prompt. So, if you ask a simple question, you'll probably get the faster, more efficient model. But if you ask it to write a complex piece of code, it'll switch to the more powerful "thinking" model. Honestly, I'm a little torn on this. On one hand, it's cool that it's all automated. On the other, I kinda liked having the control to choose the model myself.

Let's Talk About a Few Things

Now, let's get into the nitty-gritty. How does GPT-5 actually perform compared to GPT-4o? I've tested it across a few different areas, & here's what I've found.

Writing: From Pretty Good to Pretty Damn Impressive

GPT-4o was no slouch when it came to writing. It could crank out blog posts, emails, & even decent poetry. But GPT-5 is on another level. The writing is more nuanced, more creative, & just... better.

I gave both models the same prompt: "write a short poem that makes an emotional impact. A widow in Kyoto keeps finding her late husband's socks in strange places."

Here's what GPT-4o came up with:

In the cupboard with the rice bowls — one, folded neat, like he'd just stepped out. Another behind the garden broom, dusted with plum blossoms & doubt.
On the tatami, soft & alone, a heel turned toward the temple bell. One in the kotatsu, warm as breath — she weeps & doesn't tell.
The washer's empty. Always is. But socks keep surfacing, sly & kind. She never asks how they return. She only hopes he doesn't mind.

Not bad, right? It's a solid poem. But then I gave the same prompt to GPT-5, & this is what it produced:

In the tea tin, a single sock, smelling faintly of camphor & rain. Another waits by the geta rack, curled like a sleeping koi.
She lifts the futon—two more, folded into the quiet where his feet once warmed her. Kyoto's bell rolls evening down the hill.
She hangs them on the bamboo pole, black flags of a country that no longer exists, & counts them as if counting blessings, or breaths, or steps to a door that opens only when you stop knocking.

See the difference? GPT-5's version is just so much more evocative. The imagery is stronger, the metaphors are more striking, & the whole thing just has more emotional weight. It's like the difference between a student film & a critically acclaimed movie.

Coding: A Developer's New Best Friend

I'm not a professional coder, but I dabble. & I've gotta say, GPT-5 is a game-changer for anyone who writes code. It's WAY better at understanding complex requests & generating clean, efficient code. It's also much better at debugging. I've thrown some gnarly code at it, & it's been able to spot the errors & suggest fixes with impressive accuracy.

One of the coolest things is its ability to create entire applications from a single prompt. I saw an example where someone asked it to create a "Jumping Ball Runner" game, & it spat out a fully functional, colorful game with parallax scrolling backgrounds & funny sound effects. That's just insane.

For businesses, this is HUGE. Think about how much time & money you could save on development. Instead of hiring a team of developers, you could just have one person with a good idea & GPT-5. Of course, it's not going to replace human developers entirely. You still need someone with the expertise to guide the AI & polish the final product. But it's definitely going to change the way we build software.

Factual Accuracy: Fewer Hallucinations, More Trust

One of the biggest problems with previous AI models was their tendency to "hallucinate" or make stuff up. GPT-4o was better about this than its predecessors, but it still wasn't perfect. GPT-5, on the other hand, is MUCH more reliable. OpenAI claims that it's 45% less likely to contain a factual error than GPT-4o, & from my experience, that feels about right.

This is a massive deal, especially for businesses that are using AI for customer-facing applications. If you're using a chatbot on your website to answer customer questions, you need to be able to trust that it's giving them accurate information.

This is where a tool like Arsturn comes in. Arsturn helps businesses create custom AI chatbots that are trained on their own data. This means you can be sure that your chatbot is providing accurate, on-brand information to your customers. & with the power of GPT-5 behind it, that chatbot is going to be smarter & more capable than ever before. You can build a no-code AI chatbot that provides instant customer support, answers questions, & engages with website visitors 24/7. It's a pretty cool way to leverage the power of AI to improve your customer service.

The Elephant in the Room: The "Personality" Problem

Now, it's not all sunshine & roses. There's been a lot of talk about GPT-5's "personality" or lack thereof. A lot of users are saying that the new model feels more sterile & formal than GPT-4o. They miss the friendly, conversational tone of the old model.

& you know what? I get it. There were times when I felt like I was having a real conversation with GPT-4o. It had a certain... spark. GPT-5, on the other hand, can sometimes feel a bit like a know-it-all professor. It's brilliant, but it's not exactly warm & fuzzy.

OpenAI's CEO, Sam Altman, has actually acknowledged this feedback & said that they're working on making the model "warmer." So, hopefully, we'll see some improvements in this area soon. In the meantime, it's a good reminder that as powerful as these AI models are, they're still not human.

So, What's the Verdict?

Overall, I'm incredibly impressed with GPT-5. It's a significant step up from GPT-4o in almost every way. The writing is better, the coding is better, & it's more reliable. It's not perfect, of course. The personality issue is a real thing, & I'm sure we'll see other quirks & limitations as we spend more time with it.

But here's the thing: this is just the beginning. The pace of AI development is so fast that by the time we've all gotten used to GPT-5, there will probably be a GPT-6 on the horizon. & that's what's so exciting about this field. We're living in a time of incredible innovation, & we're all getting a front-row seat to the future.

For businesses, the implications are massive. AI is no longer a futuristic pipe dream. It's a real, practical tool that can help you improve your products, streamline your operations, & connect with your customers in new & powerful ways. & with platforms like Arsturn, it's easier than ever to get started. You can build a conversational AI platform that helps you build meaningful connections with your audience through personalized chatbots. It's a great way to boost conversions & provide a better customer experience.

I hope this was helpful. Let me know what you think of GPT-5 in the comments below. I'd love to hear your thoughts.