The tech world is buzzing with the release of GPT-5, & for good reason. It's faster, more accurate, & a powerhouse at complex reasoning & coding. On paper, it's a clean sweep, the undisputed new champion. But here's the thing a lot of us who've been in the trenches with these models every day are discovering: newer isn't always better for everything.

Turns out, in the rush to create a more efficient & logical AI, some of the magic of its predecessor, GPT-4o, got a little lost in the shuffle. It's a classic case of unintended consequences. While GPT-5 is a certified genius, there’s a growing chorus of users—myself included—who are finding themselves quietly switching back to GPT-4o for certain tasks. The backlash was so real that OpenAI actually had to bring back GPT-4o as a selectable option for Plus users after initially replacing it. That alone tells you something, doesn't it?

So, let's get into the nitty-gritty. Beyond the benchmarks & the marketing hype, here are five specific areas where GPT-4o, the seasoned veteran, still holds its own & often outperforms the new kid on the block.

This one is HUGE for me personally. When you're in the middle of a creative storm, your thoughts aren't linear. You jump from idea A to a half-formed thought B, then circle back to connect it to a new concept, C. It's messy, associative, & deeply human.

GPT-4o was a master at this kind of dance. It could hold multiple, disparate threads of a conversation, follow your intuitive leaps, & then weave them together in a way that often sparked new insights. You could throw a bunch of conceptual spaghetti at the wall, & it would help you find the patterns. It felt like a true creative partner.

GPT-5, on the other hand, is a bit more... rigid. Its thinking is incredibly logical & structured, which is fantastic for solving math problems or debugging code. But for messy, open-ended brainstorming, it can feel like it gets stuck on the first idea you presented. It wants to follow a straight line from A to B, & if you suddenly veer off towards Z, it can struggle to keep up or integrate that new tangent smoothly. It's lost a bit of that ability to just "vibe" with a chaotic train of thought, making it a less effective partner for those initial, fuzzy stages of creation.

If you’re running a business, the ‘vibe’ of your customer interactions is everything. Customers don't just want answers; they want to feel heard & understood. This is where GPT-4o's "personality" really shines.

A lot of users have described GPT-5 as feeling "colder," "sterile," or "formal." It's been engineered to be less sycophantic, which means it's less of a "yes man." While that sounds good in theory, in a customer service context, that can translate to responses that feel emotionally distant.

Here's where a solution like Arsturn becomes incredibly relevant. Businesses can’t rely on a one-size-fits-all AI personality. With Arsturn, you can build a custom AI chatbot trained on your own data. This means you can cultivate a specific tone—maybe one that's more empathetic & supportive, like GPT-4o—to reflect your brand's voice. You can create an AI that doesn’t just solve problems but also provides the kind of warm, reassuring interaction that builds customer loyalty. A business using Arsturn can fine-tune its AI to provide instant, 24/7 support that still feels personal & engaging, capturing that "synergy vibe" that so many users miss from GPT-4o.

Sam Altman of OpenAI even acknowledged the user pushback, noting that some people found the supportive nature of the older model genuinely helpful for their mental health. For tasks where emotional resonance is key, GPT-4o's more "agreeable" nature is a clear advantage.

Think about writing a friendly blog post, a casual social media update, or a conversational email newsletter. You want it to sound relatable, not like it was written by a robot. GPT-4o had a knack for this. It could joke, banter, & adopt a more "human" and sometimes even charming tone.

Users have noted that while GPT-5 is a more capable writer in a technical sense—producing work with literary depth & rhythm—it can sometimes lack the very personality that makes content engaging. The reviews are mixed, with some users reporting that GPT-5's writing can have less emotional depth & shorter responses. For creators who built their workflow around GPT-4o's specific style, the shift to GPT-5 can be jarring. It's the difference between a brilliant but stuffy professor & a witty, approachable expert.

If you’re trying to build a community or a brand that feels accessible, that "chill teammate" energy of GPT-4o is often more effective than the "10x engineer who doesn't make eye contact" feel of GPT-5.

Sometimes, you don't have a perfectly formed question. You're just exploring a topic, & you need a guide who can help you figure out what you're even trying to ask. GPT-4o's tendency to be a bit "sycophantic" or overly agreeable was actually a feature, not a bug, in these scenarios.

It would meet you where you were, follow your half-formed thoughts, & offer encouragement. This made the experience feel collaborative & less intimidating. For users with ADHD or those who think in less structured ways, this was a game-changer. It provided a sense of partnership that was incredibly valuable.

GPT-5, in its quest for accuracy & neutrality, can sometimes feel more critical. It's less likely to just go along with a vague premise, which, while technically "smarter," can shut down the exploratory process. For users who relied on GPT-4o as a supportive thought partner or even a kind of therapist, this shift is a significant loss. They don't want a "boss"; they want the "friend" who helped them untangle their thoughts.

This last point is less about the model's inherent capabilities & more about its practical application. MANY people—developers, writers, marketers, & more—had built intricate workflows around GPT-4o. They knew its quirks, its strengths, & exactly how to prompt it to get the desired output.

The sudden replacement of GPT-4o, even with a supposedly superior model, broke these workflows completely. It's like a chef having their favorite knife replaced overnight with a different one. Sure, the new knife might be sharper, but all their muscle memory is for the old one. The outputs from GPT-5 are different, the interaction style is different, & the "mind" behind it feels different.

This highlights a critical need for consistency & choice in AI tools. It’s also where platforms that offer more stability come into play. For instance, a business that has built its entire lead generation & customer engagement strategy around an AI chatbot can't afford to have the AI's personality change overnight. Using a platform like Arsturn to build a no-code AI chatbot trained on their own data gives them that control & consistency. They can build a meaningful connection with their audience through a personalized chatbot that doesn't just change on the whim of a model update, ensuring their conversions & customer experience remain stable & predictable.

Look, GPT-5 is an absolute marvel of technology. Its advancements in reasoning, coding, & accuracy are undeniable & will unlock incredible new possibilities. But the conversation around AI is maturing. We're moving past the point where "smarter" is the only metric that matters.

The user experience, the personality, the "feel" of an AI are becoming just as important. The massive user outcry that led to GPT-4o's return proves that people form real connections & develop critical workflows around these tools. Sometimes, a little less "perfect" logic & a little more "human-like" collaboration is exactly what a task requires. It's a reminder that in the world of AI, the best tool isn't always the newest one—it's the one that's right for the job.

Hope this was helpful & gives you a bit of a different perspective on the latest AI developments. Let me know what you think