OpenAI has officially launched GPT-5 – the next generation of its most famous chatbot AI – and made it the default ChatGPT model for all users, even free ones. If you’ve ever felt like ChatGPT was always a bit too “robotic” or that you’re drowning in models with names that sound like processor codes, then GPT-5 will be a real simplification for you.
ChatGPT 5 is not just a cosmetic retake exam. It is about major renovation, which combines speed, increased intelligence, new tricks, and – yes – even the ability to choose his personality.
Tests show that ChatGPT 5 30–50 % faster than GPT-4 on simple questions. If you want to know “How many rings does Saturn have?”, you’ll get the answer in three seconds, while the old model took twice as long. For light tasks, this wasn’t the biggest problem with previous generations, but the responsiveness feels much more “instant” now.
All brains in one model – ChatGPT 5
Perhaps even more important news: we said goodbye to the model selector. Previously, you had a whole range of options available – GPT-4, 4.1, 4.5, 4 Mini High … and each one was a little better at one thing, worse at another. The average user felt like in a restaurant with too many similar dishes. But the advanced user knew exactly what to expect from a particular model and therefore used the one that gave him the best results for a particular task. Now – ChatGPT 5 does this itself, which is extremely wrong and a wrong decision, because the user – that is, the advanced one cannot choose any of the old models.
ChatGPT 5 makes this simple: a single model It combines the capabilities of all specialized versions and decides when to use the “easy” mode for a quick question and when to slow down and use more “brain power” for complex tasks. The result? When creating a game of Tetris from a single sentence, GPT-5 not only sleeps the basic version, but intuitively adds scores, levels, and displays the next pieces – all without additional instructions.
ChatGPT 5 personalities: from cynic to nerd
A new feature you'll either love or ignore: ChatGPT can now chat in your chosen personalitiesAvailable are:
- Cynic – dry humor, sarcasm, directness. Like a friend who makes fun of you but helps you anyway.
- A robot – no emotions, no twists. German precision in digital form.
- Listener – warm, relaxed, a little therapeutic. Like a bartender who understands you.
- Nerd – curious, playful, excited about any new information.
If this sounds like too much drama to you, you can always choose the default style.
Vibe coding 2.0
For developers and the curious: GPT-5 is even better at vibe coding – writing code from natural language. You describe what you want, and the AI in Canvas creates a working app or website that you can test immediately. Previously, specialized models (like 4 Mini High) were faster, but now GPT-5 combines speed with more complexity. For example, when playing chess with Pokémon, it not only replaces the pieces, but also adds rules, highlights of possible moves, and visual details that make the whole thing look “premium” – and work without stumbling.
More appearance customizations
ChatGPT 5 can now be customized to your style – from choosing a color accent for message bubbles to a voice button and highlighted text. It’s a small thing, but it feels like AI is sitting at a table in your personal “UI lounge.”
Advanced voice mode that adapts
Paid users get almost unlimited access to advanced voice mode, which not only sounds more natural, but you can adjust its tone – from a calm radio host to an enthusiastic friend after three coffees. The standard voice mode is gone, and free users get a limited number of “talking hours”. Bonus: it also works with Custom GPTs.
Gmail and Google Calendar integration
GPT-5 can now be connected to Gmail and Google Calendar, meaning the AI can:
- organizes your day,
- warns you about unread emails,
- coordinate meetings,
- and all this without having ten open tabs.
The feature will initially be available to Pro users, with others coming later.
Weak points: facts and pictures
Despite the promises of OpenAI GPT-5 still likes to hallucinate – invents products that don't exist and presents them with complete confidence. At the same time, it seems to be worse in terms of text, at least when it comes to helping write articles, where it behaves very much like the GPT 4 model and, for example, worse than GPT 5.
The progress in image generation is not as obvious – sometimes older models still offer better composition and more of a “wow” effect. It feels like they’ve optimized the performance and tried to save money on many fronts rather than giving it their all.
Writing with more "fine feeling" - really?! We'll see.
When it comes to writing, the difference is clear: GPT-5 is more natural, uses analogies, builds a narrative arc, and sometimes even adds directorial notes for filming. But, for all those of us who write with it (or have been helping each other for years) and knew in which model, we get what, and where it is worse! It seems that it understands the instructions of custom “GPTs” less and follows them generally worse. Less precise and that it writes without a character that you could “tame” to your own taste.
Final impression: better in some places, worse for professionals in certain areas
GPT-5 is a significant upgrade: faster, more intuitive, and easier to use. Unifying models means less clicking and more action, and personalities and customizations add some human touch. It’s still not perfect—facts sometimes escape into the realm of imagination, visual creativity isn’t always at its peak—but as an everyday assistant, GPT-5 is a step closer to an AI you’ll actually want to use.
More brains, fewer buttons, more fun – with a little skepticism in the package.
