If the question is which of the great technology is winning the AI career, the answer is: None
Who is winning the AI race? At this point we should have it more or less clear. We had it when Microsoft and Intel were profiled as the dominators of the PC world or when Apple and Google triumphed with their smartphones. But with AI something curious happens: Things are very even. Of, done, too much. Of course, in the field of Openai popularity, it is the most prominent with chatgpt. Recently the company presumed to be touching The 700 million active weekly users, a really remarkable figure that leaves its competitors behind. However, that metric is not definitive, especially when we have a great unknown to elucidate: what is the best AI model? It is impossible to know today what is the best model of AI No one can give a clear answer to that question. Neither the companies, which continuously breastfeed with their new versions, nor the benchmarks, which They have become a useful but imperfect tool when evaluating the quality of these models. In Polymarket people believed that the best AI model at the end of August was going to be OpenAi. After leaving GPT-5, the perception changed. That the answer to that question is difficult Polymarket demonstrates itthis unique prediction platform in which users bet on a result and also pay it by voting for one or another conclusion. To the question of “What company does the best model of AI have at the end of August? Everything seemed to smile at Openai, but after the launch of GPT-5, the Batacazo: now the clear favorite is Gemini, the Google model, with OpenAI collapsed to 16% of the votes and even more behind Grok (XAI) with 6.3% of votes and Claude (Anthropic) with a very low (in my opinion) 1.5%. It is not that Polymarket is an especially reliable indicator of this (or anything), but it makes it clear that The public perception of these models can be very different of their real behavior in things like their number of users-here OpenAi tombs their competitors-or their performance performance such as Arc-Agi 2 (where Grok 4 wins everyone, including GPT-5). In the Benchmark of abstract thought ARC-AGI 2, Grok 4 is well above its competitors. Included GPT-5, which exceeds Claude Opus 4. Source: ARC-AGI. And that makes us even clearer what are the two great reasons why it is a real problem to know which model of AI is winning this career. The first, that these tests are often very specific and concrete, and focus on evaluating aspects such as the ability to program or solve mathematical problems of these models. And the second, that The models do not stop improving and to overcome what his rivals had achieved a few days, weeks or months before. We do not stop seeing how new versions of the models are (logically) something better programming, generating text or images or solving certain types of problems, but there is no consensual or definitive form of saying “this model is better.” As we have seen, each user also has their own personal perception (Hello, Polymarket) when using them. Some prefer Claude to program, other chatgpt for generic questions, other Gemini to talk about diverse topics and to learn, for example. And none seems to be the final model “for everything.” In a recent scientific study, researcher Steve Hsu concluded that the path followed by the current generative models will not lead to AGI. Neither now, nor ever. That leads us to a reflection: that this general artificial intelligence (AGI) is far from arriving. These systems, which are supposed to be overcome in all areas, are not even remotely close to doing so, and They continue to make mistakes even when for example GPT-5 seems to have significantly mitigate the problem of hallucinations. Analysts like Gary Marcus They reminded us These days that have been saying the same thing about 30 years: that with this type of climbing techniques We are not going to get to an AGI And that the road has to be another. And that leaves us some interesting ideas. David Sacks – Paypal Cocfounder, founder of Yammer, investor— analyzed The situation of this segment and raised striking conclusions. The five main companies that develop foundational models —Openai, Google, Meta, Anthropic or XAI – still do not master the market, but that is (or it can be) good news. And it is because there is neither a monopoly nor a duopoly of AI. What there is is a fierce competition not only among these five North American companies, but between them and all their Chinese competitors, to which a lot of startups are added that have no resources to work on foundational models – careful – and instead they try to solve another great question: what is the Killer app of the AI. That’s where there are great opportunities for these startups, which can solve success cases in which AI can really be a disruption for an industry. It is for example what emerging companies have done such as cursor or Windsurf, which have opted for the vibe coding boom and are capturing a lot of interest among the developer segment. In fact, every time we see how even the greats of AI presume that their new models program especially well or perhaps are more oriented than ever to solve mathematical problems. GPT-5 precisely use those two arguments to declare themselves better than the competition, and although some benchmarks prove them right, the perception of the users will determine whether they meet expectations or not. But there is also that great debate between the proprietary models (such as GPT-5) and the Open Source models. As Sacks says, the fact that Open Source models They can offer 80-90% of the capacity with a cost of 10-20% of foundational models is sensational for certain users. Specifically, for those who prioritize customization, control and cost savings on the use of foundational models. China goes for all with that philosophy, although curiously it was … Read more