GPT-5 promised a lot: its biggest secret is that it is a Trojan horse for everyone to use AI.

What if OpenAI's biggest leap forward isn't a technical breakthrough, but a shift in how we use artificial intelligence in our everyday lives? OpenAI unveiled its long-awaited GPT-5 model this Thursday , the one that was said to have given Sam Altman the creeps . It may not be that big of a leap, but it's certainly going to make AI usable by anyone, even the most inexperienced.
GPT-5 not only speaks better, thinks more, and hallucinates less. It also decides for you how to respond, and for the first time, it's available to everyone. You no longer need to pay to access high-level AI. Just open ChatGPT, because starting this week, all users, even free ones, will have access to it.
But, amid measurable improvements and promises of accessibility, something has changed in tone. GPT-5 hasn't been received with the same euphoria as GPT-4. Is it because we've reached a threshold of sophistication where the leaps are no longer as noticeable? Or because AI, as it becomes more useful and ubiquitous, is also becoming less surprising?
Accessibility as a flag: GPT-5 is coming to everyoneGPT-5 represents, above all, a redesign of the AI user experience. One of the most notable changes is not in what the model can do , but in how it integrates into our routines. The new "auto-routing" system automatically decides whether to provide a quick response or take a moment to "think." This dynamic choice eliminates the need for the user to manually choose between different versions of the model or activate features like "thinking mode."
Better integration with platforms like GmailAdditionally, GPT-5 has been more deeply integrated with everyday applications like Gmail, Google Calendar, and even task managers . This makes ChatGPT closer to a real assistant than a simple chatbot. It can help compose emails, schedule meetings, or generate reports without the user having to leave the interface.
Another key move has been to open the model to free users, breaking with a tradition maintained since GPT-4. According to OpenAI, the goal is clear: to fulfill its founding mission of making advanced artificial intelligence accessible to as many people as possible . In the words of Nick Turley, VP of ChatGPT: "This is just another way of life."
The result is a model that not only can do more, but also reaches more people.
Technical improvements, but no quantum leapFrom a technical perspective, GPT-5 shows solid progress, but not as spectacular as one might expect. In coding, for example, it excels with 74.9% in the SWE-bench Verified test, beating Claude Opus 4.1 and Gemini 2.5 Pro . In doctoral-level science tasks, such as GPQA Diamond, its performance is also notable, with 89.4%.
In the health field, GPT-5 only "hallucinates" in 1.6% of cases when its deep reasoning ability is activated, a significant improvement over previous models. It also shows greater proactivity in interpreting medical results and alerting about possible symptoms, although it does not, of course, replace a healthcare professional.
However, its performance in simulated online tasks, such as web browsing or shopping, was uneven. In some cases, rival models like the Claude or Grok slightly outperformed it. Also in the "Humanity's Last Exam" test, the GPT-5 Pro fell behind the Grok 4 Heavy.
Does this mean GPT-5 isn't better? Not necessarily. The improvements seem to have focused on more subtle aspects, such as error reduction, natural responses, and a greater ability to identify malicious requests. All of this makes the model more reliable and useful, although less conspicuous.
User experience: more personalized, less frictionOne of GPT-5's most visible innovations lies in its interaction design. OpenAI has introduced four new preconfigured "personalities" —Cynic, Robot, Listener, and Nerd—that alter the way ChatGPT responds without requiring specific instructions.
This type of lightweight customization, combined with app integration and the elimination of complex menus, creates a ChatGPT closer to a "personal digital agent." We're not just talking about better responses, but a smoother, less frictionless experience.
There are also new features for paying users. Subscribers to the Plus plan ($20/month) have greater access to the base model, while those on the Pro plan ($200/month) can use GPT-5 Pro with unlimited access. The latter model utilizes more computing resources, resulting in more precise responses tailored to complex tasks. Businesses, educational institutions, and developers will also be able to integrate GPT-5 into their platforms via the API, which is available in mini, standard, and nano versions.
Comparison with other models: The best in the world?During the presentation, Sam Altman didn't hold back: he called GPT-5 "the best model in the world." Still, the data presented invites a more nuanced interpretation. In several benchmarks, GPT-5 clearly prevails. In others, it barely outperforms its rivals or even falls behind.
However, its overall behavior shows a more stable AI , less prone to serious errors, and with a greater ability to understand complex contexts. The improvement in "taste" or creative judgment is difficult to measure, but it has been highlighted by OpenAI executives themselves, who describe the model's "vibes" as particularly apt for design or writing tasks.
Security issues have also been mitigated. GPT-5 is less prone to manipulative behavior and better at distinguishing between dangerous questions and legitimate requests. This not only improves system security but also the user experience.
A pragmatic evolution… but not transformativePerhaps the reason GPT-5 has left some observers somewhat cold isn't its lack of advancements, but rather the absence of a disruptive narrative. It didn't have the emotional or conceptual impact that GPT-4 had. Instead, we find a more polished, useful, and accessible tool. A quiet advance, yes, but a profound one.
The real innovation of GPT-5 may not be in its numbers, but in how it begins to slip into the cracks of everyday life. We no longer need to understand what a "reasoning model" is or choose between "GPT-4o" or "o3." GPT-5 decides for us. It listens to us, helps us, and anticipates.
But this accessibility comes at a price: the illusion of magic is diluted. As if we're seeing fewer fireworks because AI has now stepped down from its pedestal and quietly settled on everyone's desktop.
eleconomista