openai

Microsoft opted everything to OpenAi to win the AI ​​race. Start realizing your mistake

by

In April 2019, a small group of Openai engineers flew to Seattle to make a demo of a GPT-2 Supervitaminated Version A Bill Gates. Microsoft’s co -founder was impressed, and made clear his intention to invest in the company led by an almost absolutely unknown Sam Altman. In Redmond, Gates’ opinion took into account, and shortly after the events precipitated. The company’s cto, Kevin Scott, wrote an email Shortly after, in June, Nadella already Gates to warn of the danger of being behind. He was “very, very worried.” Nadella agreed. A month later, on July 22, 2019, Microsoft advertisement a Investment of 1,000 million dollars In OpenAi. That seemed love at first glance. That money promoted Openai’s efforts, which continued to develop its models and that in August 2020 launched an amazing GPT-3 But without opening it to the general public. Two years later, the bombing: OpenAi announced chatgpt without knowing that this would detonate the current AI fever. This divorce can be very expensive to Microsoft The initial impact was amazing, and Chatgpt became On the fastest platform he had grown up in history of the Internet. In two months he had already attracted 100 million users, and Microsoft, wanting to take another step in that idyllic relationship with Openai, redoubled its commitment to the company. Or multiplied it, because in January 2023 advertisement a multimillionaire investment that is estimated – although the figure was never detailed— at 10,000 million dollars. That investment It seemed to be perfect For both companies. Among other things, Openai obtained access to Microsoft’s cloud infrastructure to train their models and offer them to the general public (inference). Meanwhile, Microsoft gained priority access to OpenAi models, which it could sell as if they were his. That is just what they did, first with GPT-3 as a Github Copilot baseand then Freaking of other “co -drivers” that were nothing more than a chatgpt rehash. And there began the problems for Microsoft. Above all, because little by little that theoretically idyllic relationship He began teaching his seams. The convenience marriage was no longer so satisfactory for both parties. On the one hand, Openai kept asking for more and more money and better conditions when using the Microsoft computing cloud. On the other, Microsoft, OPENAI’s totally slave For his AI options, he began to look for alternatives. Both decided to look for alternative plans. Openai searched Change girlfriend and of Alliesand After the rumors In April 2025 he announced the greatest financing round in history, which amounted to 40,000 million dollars. For his part, in Microsoft, knowing their Openai absolute dependencethey began to move in March 2024, when they created their own division of AI and put in front of it To Mustafa Suleymanco -founder of Inflection AI and before Deepmind. The objective: to develop their own foundational models to avoid being chained to OpenAI. How was the thing? For now, OpenAi very well. To Microsoft, not so much. The company led by Sam Altman has not stopped growing in users and In income. At the moment they are still insufficient to make it profitable, but there is a clear thing: Nowadays Chatgpt is what Google went to searches. There are options, yes, but for the vast majority of users, (almost) do not count. I may launch of GPT-5 has been disappointingbut still Altman’s strategy to sell promises and Hype -with the Stargate megaproject in front – it works. But for Nadella and yours are much more complicated. Despite Copilot’s absolute integration in all business areas, Microsoft is almost an “coupled” in the world of AI. A pay. Your own modelsPhi-3 and Phi-4 are interesting for their “Edge AI” (artificial intelligence that runs at home, as in our mobiles), but their performance and capacity has made them almost a laboratory experiment. The gigantic infrastructure of Azure is his great asset to make them indirect leaders of AI, but Suleyman’s leadership It is committed to the simple reason that the results of the Microsoft strategy are not especially visible. It doesn’t matter if Microsoft has a powerful Trojan horse in Windows to infiltrate AI solutions in the company: who are winning that battle They are Anthropic and, of course, OpenAi. Meanwhile, Suleyman himself published on his blog An article warning of the dangers of treating AI as a person and ensuring that we are close to seeing a “apparently conscious” that can aggravate that problem. It is a valid and important argument, but it leaves Microsoft just like it was: without its own models and without changes in a strategy of doubtful success. In spite of everything, Microsoft also has its badges For Microsoft the only consolation is that its gigantic investment in OpenAi gives you the right to participate in the benefits of that company. In fact, the tense current relationship between the two adds to Openai’s intentions of becoming a profit company. That has implications for the future of the relationship, and In The Information indicated that Altman has proposed yield to Microsoft 33% of the company But renouncing future benefits. Participation is enormous and very juicy for Redmond’s, especially now that Openai is valued in 300,000 million dollarsbut there is much more at stake. In fact, in that potential divorce the one that seems to be winning is Openai, which has some of the best foundational and popular models in the world (GPT-4O, GPT-5), and who is the beautiful girl in the market: everyone wants to stick to it. Meanwhile, Microsoft is staying absolutely behind in the market, at least compared to its rivals. Let’s see: Google: It has invested a lot and well in own models (Gemini, Deepmind) and has a gigantic infrastructure, software and data. Not to mention Android and his search engine, AI entrance door for billions of users. XAI: Although less remarkable, the startup created by Elon Musk has managed to stand up with a Grok with a very different approach –zero censorship– which has managed to integrate with some … Read more

OpenAi already enters 1,000 million dollars per month. They are crumbs compared to what you need to be profitable

by

The launch of GPT-5 It has been rather disappointing for users at street level and has arrived accompanied by some problemssomething that has recognized the Sam AltmanCEO of OpenAi. However, it is sweeping a key sector: Companiesfor the quality of its reasoning models and its price. Not being a company that quotes in the stock market, the doubt is how much success translates into large income, and Sarah Friar, Financial Director of OpenAi has given a key fact: OpenAi has entered 1,000 million dollars in a month (July) for the first time in its history. Results as good as insufficient. Openai’s reality is that it is still early to celebrate a figure like that announced by Friar, especially when the executive recognized that artificial intelligence “right now is voracious in GPU and in a computational power.” In that sense, he affirmed that the biggest problem they face is “being constantly under computational capacity.” That is, the demand for the use of GPUs of its models is greater than the resources they have to cover it. The plan to cover your needs is clear, and Friar recalled it: “That is why we launch Stargate”, in reference to that Historical project In which partners such as Microsoft, Oracle, Nvidia and ARM participate, and in which they expect to spend 500,000 million. They expect, because There is still no single contract signedand find more sources of financing It is a huge challenge. With this financial context, the 1,000 million dollars that OpenAi already enters a month are totally insufficient. They look for solutions, but more complications arise along the way. The reality of artificial intelligence according to Sam Altman is that his demand will continue to grow, their training needs will continue to grow and spend “probably more aggressively than any company has done in something for progress.” Given this need, OpenAi seeks solutions, such as selling 6,000 million in shares with an assessment of 500,000 million dollars, after raise 40,000 million in a financing round at the end of Marchwith an assessment of 300,000 million. 18,000 million were going to dedicate themselves to the Stargate projector. The problem is that to get the last 30,000 million of that round, Openai had to become a “For-Profit” company, a path that He has recently abandoned. A problem structure of expenses. As much as Openai continues to have the confidence of large investors such as Softbank or Microsoft, the problem is that the company’s expenses are huge. Not being a quoted company we cannot know official figures, but estimates point to 8,000 million operating expenses per year (They do not count investments, infrastructure or other financial obligations). Even if it continues to enter the current rhythm (1 billion per month), in 2025 it would not go from 12,000 (the estimate they handled for this year, According to The Information). That is, only their operating expenses “eat” 66% of the operating expenses, which do not contemplate their higher volume investments. The cost scheme is very complicated, and Sam Altman came to affirm that They lost money even with chatgpt prothe subscription of 229 euros per month. And the profitability for when. According to An internal studyOpenai will lose 44,000 million dollars between 2023 and 2028, and 14,000 million only in 2026, the triple of those estimated by 2024. It will have to wait four more years, until 2029, so that the company exceeds the Break Even And I achieve The long -awaited profitability. It is something they estimate that they would get 100,000 million dollars annually. Right now, viral moments such as the generation of chatgpt ys imagesor explosion of images of Ghibli either help: “Derritate” servers with a huge cost. The good news is that Sarah Friar confirmed that they have seen acceleration in the adoption of payment subscriptions. But much remains to stop losing money. Image | Dima Solomin In Unspash, Village Global In Xataka | Google has finally revealed how much electricity and water consumes its AI. Estimates could not be more wrong

While everyone criticized GPT-5, Openai was winning the war that really matters: that of companies

by

He GPT-5 launch It has been, in broad strokes, disappointing. Openai needed this model with this model bigger in the history of AIbut we have encountered a model that improves, but not spectacularly. And yet, it is achieving something that is more important than it seems: to convince companies. Companies

GPT-5 has many versions, but Openai decides which one to use for you

by

Openai wants to simplify our lives, but maybe he has made a mistake in doing so. When you use GPT-5, you won’t know which GPT-5 version you will be using. Because although Openai has raised it as a “unified model”, it is actually using variant distress according to the scenario and, of course, according to what you pay. And not knowing what he is using at all times is a blessed? problem. Welcome to the era of the opacity of AI. GPT-5 are many GPT-5. In his “system card“Openai stands out how GPT-5 It is a “unified system” that actually consists of several versions. Some more powerful, others less, and others that reason. But unlike what happened so far-you chose if you wanted to use the lighter model or one that would think more-GPT-5 will choose for you. As? On the one hand we have the “external” names of the variants of GPT-5 and on the other the inmates. The “real -time router” will be responsible for deciding which is used at each moment. Source: latent.space A “router” that chooses for us. To know what variant you have to use, GPT-5 uses a “real-time router that quickly decides which model to use based on the type of conversation, complexity, or the need for tools.” As they point out In Latent Space The idea is remarkable, and helps to avoid That name chaos We had so far. Of course: the only way to “force” the type of model to use is to express it clearly in the prompt when using chatgpt. Thus, if we ask him something and ask that “think a lot about this”, we will activate that reasoning model so that the router does not make mistakes by choosing perhaps a less capable model. Opacity. Although OpenAi indicates that he uses these arguments to decide which GPT-5 variant USA, here we face the dichotomy of opacity. That GPT-5 chooses for us is on the one hand good and on the other bad. Simplifies the use of chatbot, no doubt, and that brings it closer to the novel public. But at the same time Advanced users steals control or those who have enough knowledge to know how to decide which version of GPT-5 would like to use. With an expected exception. We can already say goodbye to this old model selector. For better, and for worse. Beware of simplifying. The simplification problem of complex technologies is that it can cause unwanted side effects. For example, that users do not understand the differences between the models. That leads to reducing its ability to make informed decisions about which tool is more appropriate for each specific context. Openai raises something like a “technological infantilization”: do not worry, dear users, that GPT-5 will already do everything for you. Even choose how to work. If you pay, choose. If you are one of those advanced users who know how Chatgpt Pro (200 dollars/month) In order to activate any of the different internal models of GPT-5 Correspondence between the “old models” and the “new models.” Source: OpenAi How are they compared to previous models. The arrival of this new family of models assumes that now each of the above It has its replacement with one of the Variants of GPT-5. Thus, the old GPT-4O would be the new GPT-5 Main, while GPT-4O-mini will be GPT-5-Main-Mini and O3 will be GPT-5-Thinking. And so with the rest: Goodbye, old GPTS. As explained in the version notes Published in OpenAI, the arrival of GPT-5 supposes goodbye to the models that were used so far in Chatgpt. Chatgpt Pro subscribers They can continue to have access To these ancient models, at least for now, but for other users the option is clear: the pre-GPT-5 models are retired, and among them: GPT-4O (will continue to be used for voice mode) GPT-4.1 GPT-4.5 GPT-4.1-mini O4-mini O4-mini-High O3 O3-PRO GPT-5 should explain your decisions. Although the idea of having a unified model is not bad, that lack of control and information about the way of functioning of GPT-5 is worrying. One of the solutions is to introduce that selection of ways (available in Chatgpt Pro), but another is even simpler: Chatgpt would be enough to indicate which version of the model has used at all times so that we know what to stick to … and if we can force that in such questions a more capable (or less) model is used. In Xataka | The AI has so far a nice accessory: with GPT-5, OpenAi wants to place it in the center of all things

A scheme to understand the new Openai in chatgpt

by

GPT -5 is here. If you have reached this article, it is probably for one of two reasons: o You are interested in knowing what it contributes The new OpenAI modelor you have encountered an unexpected change in your application of Chatgpt. The model selector, in general, has disappeared, and with it, part of the control that users had. Between what GPT -5 promises and what is now out of reach without paying, the question is served: is it worth subscribing? What has really changed in Chatgpt’s experience? GPT -5 decides on your own how to answer GPT-5 does not work as the previous models. Now chatgpt incorporates a system that Decides automatically How to answer. If it detects that a task is simple, prioritizes speed. If you consider that you need more analysis, it activates its mode of reasoning. Openai describes it as a system that learns from previous patterns and adjusts its behavior based on the context. The result is a more uniform experience, but also more closed. The only exception is the users of the Pro and Team plans, which can continue to manually access the different variants of GPT -5. What do you get according to what you pay The result you get when using GPT -5 in chatgpt depends directly on the plan you have. It is not just a matter of speed or quality of the answers: daily limits changethe available tools and the type of reasoning that the model can apply. And in some cases, even what you see on the screen. OpenAI has designed three access levels for the general public: the free plan, The Plus Plan of 23 euros per month and The Pro Plan of 229 euros per month. Each offers a different experience. Knowing what changes between them is key to understanding why your chatgpt does not behave like that of other users. The following table summarizes what each of the three main levels includes: The free, the plus and the pro. Some differences are subtle – like automatic degradation to a lighter model – and others are much more visible, such as access to advanced tools or GPT -5 Thinking Pro mode. Free Plus pro Base model GPT –5 GPT –5 GPT –5 Allowed messages 10 every 5 hours 80 every 3 hours Unlimited What happens when exceeding the limit? Change to GPT-5-mini Change to GPT-5-mini – GPT -5 Thinking mode (automatic) 1 time a day No restrictions No restrictions GPT -5 Thinking mode (manual) Not available 200 weekly messages No restrictions Thinking Pro Not available Not available Available Image generation Around 3 a day Higher quota Unlimited Advanced voice mode Not available Available Available MODEL SELECTOR MANUAL Not available Only between GPT-5 and GPT-5-Thinking GPT-5 and GPT-5-Thinking. Old models for limited time Monthly price 0 euros 23 euros (includes 3.99 euros of VAT) 229 euros (includes 3.99 euros of VAT) When you use free chatpt you start with GPT -5 in normal mode, but the accountant does not stop: 10 messages every 5 hours. Upon reaching the tenth, the conversation changes to a distilled copy called GPT-5-mini. In addition, the Free Plan only allows a daily message with the way of deep reasoning (GPT -5 Thinking). What really contributes to pay The free plan gives access to GPT -5, but with cuts that appear quickly. Paying solves that initial limit, but also changes the experience significantly. With plus, you can stay in the full model for much longeruse advanced tools and manually force the way of deep reasoning, although with a weekly limit. Pro goes further: it eliminates virtually all restrictions. There is only a “limit” that seeks to avoid abusive use. The most important thing about the Pro Plan is not alone in the number of messages. What it offers is GPT -5 Thinking Pro, a version of the model that uses more computing to think, compare possible answers and reduce errors. This extended reasoning is not available in Free or Plus. The difference is especially noticeable in complex tasks: code, long summaries, reporting reports, technical analysis or complex comparative. For those who use chatgpt as a work tool – and not only for specific consultations – this subscription can be very valuable. Only those who pay more retain total control The new GPT -5 system eliminates the need to choose models … unless you are part of the few who can still do it. Only users of the Pro, Team and Enterprise plans retain access to the complete models selector, which allows manually changing between GPT -5, GPT -5 Thinking, and GPT -5 Thinking Pro. They can also access, for a limited time, some old models for compatibility. As we say, in the rest of the plans – Free and Plus – the experience is totally unified: the system decides for itself how to act, but we can force quick or reasoned responses through the prompt. Images | OpenAI | Omid Armin | Xataka screen capture In Xataka | The AI has so far a nice accessory: with GPT-5, OpenAi wants to place it in the center of all things

With GPT-5, Openai wants to place it at the center of all things

by

Microsoft and Apple defined the PC segment and They made him a Commodity. The computer became in something standardcommon and easily interchangeable, Without too much differentiation. Those who ended up winning that battle, however, were Microsoft and Intel, who took a good part of the benefits. Then the mobile would arrive. It is clear that those who have won this segment have been Apple (with iOS) and Google (with Android), but be careful, because those who defined it were two companies. The first, the aforementioned Apple, which also triumphed with the APP Store paradigm And it has ended up taking about half of the income. The second, a company where perhaps not many would have thought: Amazon. The electronic commerce giant would end up promoting the cloud with its Amazon Web Services, and It also ended up turning that cloud into another Commodity. And as happened with Microsoft and the PC or Apple and the mobile, Amazon ended up monopolizing a good part of the benefits of the cloud. A cloud we considerably access (for example, through the iPhone), but to which we don’t even pay too much attention: It is there and works, as before the PC and now the mobile. As explained by analyst Ben Thompson In StratecheryThey are everything Commodities. Well, there are those who expect AI to end up becoming another Commodity. That this segment for which now They compete a multitude of companies It ends up becoming (perhaps) a duopoly like the one created by Microsoft and Intel in the PC or Apple and Google on the mobile. Foundational models are the new PC/mobile, and although there are strong candidates to end up winning that race, everything is yet to be defined. AI like Commodity And that is where GPT-5 comes into play, which wants to be something like Apple’s App Store or Amazon AWS: that platform that Turn it in the company that defines the AI race and manages to win it. And that over time also influence that conversion of AI in a Commodity. With this OpenAi launch you play a lot. Sam Altman has been living in two things for years. On the one hand, of the success of Chatgpt, a remarkable product, but for the moment not revolutionary. On the other, Above all, of Hype. Of promises and moments wow like the one they gave us With GPT-4O or with them Studio Ghibli Style Images. Because we insist, Chatgpt is a good product that has proven useful as an assistant in all types of tasks, but especially in programming. Has unusual an unusual fever for the generative AI and data centersand has caused many companies to trust their future –or rather they bet– To this technology. But revolutions, for the moment, few. Comparisons with phenomena such as PC, Internet or mobile are inevitable. And in all of them discourse and evolution were similar. There were doubts and skepticism at the beginning, exaggerated bets later –Hi, bubble Puntocom– And, indeed, revolution. And it is expected that something like this passes with the pass, although in fact the most devout of this technology claim that this will go further. Others, much more skeptical, believe that This is a bubble that will end exploding … although it can end it deriving the true revolutionas happened with the Internet. Whether there is bubble and explodes as if not, GPT-5 is for Openai a key model for your future. And the paradox of this model is that the company needs the greatest jump in the history of AI Just when it seems most impossible. It is about to see how good the model is really, but the company cannot suffer Another fiasco like GPT-4.5 or your reputation will be very damaged. Will Openai achieve with GPT-5 and the AI what Microsoft, Apple or Amazon already achieved in their respective segments? That is the big question. And we will have to wait a bit to answer it. In Xataka | There are too many AI models. That raises a true death sentence for Anthropic and Claude

This is the launch of the most anticipated artificial intelligence of Openai in an increasingly fierce race

by

For months, the feeling was clear: OpenAi needed a blow to the table. GPT-4 had raised the bar, and GPT -4 -or reinforced that base with concrete improvements in performance and multimodality. But GPT –4.5 He did not convince: for many, it was a false step, With few real improvements and too many promises without compliance. Today, With GPT -5, the company seeks to recover impulse. The new model is now available and arrives in three versions. Openai has profiled GPT -5 as its most complete model to date. His training puts the focus on three great pillars: understanding more precision instructions, managing longer contexts and working hand in hand with external tools. The ambition is not less: transforming the model into a kind of digital agent that not only responds, but acts, plans and adapts. The clearest difference is how we interact with him. GPT -4 forced to choose: speed, vision or reasoning. GPT -5 breaks with that barrier. Now it is the model who decides in real time if he must respond instantly or stop to think. The result is a more fluid experience, without the need to configure modes or activate specific functions. GPT -5 is already the new default model in the free chatgpt plan. It is the first time that anyone can access an advanced reasoning system without going through a box. In payment plans, the model is deployed with more powerful versions: Chatgpt plus ($ 20 per month): access with extended limits Chatgpt Pro (200 month): unlimited access to GPT -5 Pro, a variant reinforced with more power, better answers and extended functions (voice, search, canvas, deep reasoning) Enterprise, Team and Edu plans will receive it next week as a base model. In development. Images | OpenAI In Xataka | I have tried the new OpenAI models. It has been a small odyssey with prize: I have a chatgpt at home

OpenAi has just kick the AI board. Your new model is free and can be downloaded to use from the laptop

by

Although his name suggests a certain vocation of technological opening, Openai had been moving away from that spirit for some time. From long before the launch of Chatgptin November 2022, the company did not publish an open weight model. In fact, it is the first open weight model since 2019 (GPT-2). That has just changed. This Tuesday, OpenAi has announced GPT-Ossan open weight model under Apache 2.0 license that can be downloaded for free and, in its small variant, run even in a laptop. A movement that nobody expected at all, especially since all eyes were put in Lat the imminent arrival of GPT-5. What is GPT-Oss? GPT-Oss is a language model developed by OpenAI, capable of understanding text instructions and generating answers with a high level of coherence, context and utility. It is a variant of the GPT family, the same technology that gave rise to Chatgpt, but with a key difference: its weights are open. That is, you can download it, run it on your own computer and adapt it to your needs. For the first time in years, OpenAI launches a powerful model that is not tied to its platform. You can use it without internet connection, without depending on your servers and freely to customize it. And besides, it is free. That opens the door for developers, small companies or even advanced users can experiment with AI in their own terms. What means that the weights are open? When we talk about the “pesos” of a model are open, we refer to the internal numerical values that you have learned during your training. They are, in essence, the knowledge of the model: what allows you to reason, write, translate or generate useful answers. That OpenAi has released those weights means that anyone can download them, analyze them and use them outside their servers. It is not the same as releasing the complete source code or the data with which it was trained, but it is an important step: it allows you to execute the model in own environments, adapt it to concrete tasks or integrate it into products without having to go through the Openai API. So is it a really “open” model? Now, the weights are open does not mean that the model is completely open. GPT -Oss is not Open Source strictly speaking: OpenAi has not published the data with which the model or the full code of the training process trained. What has released, as we say, is the final result, the parameters learned, under a permissive license (Apache 2.0). That is why we talk about an open weight model, not open source, According to the standards of organizations such as the Open Source Initiative (OSI). Can I use it on my computer? It depends on the model. GPT-Oss occurs in two versions: one of 120,000 million parameters and a lighter of 20,000 million. The most powerful version requires a GPU of at least 80 GB of memory (professional environments). The 20B version is much more accessible: it can be executed on computers with 16 GB of memory, which opens the door to use in personal portables and PCs. It is already compatible with tools such as LM Studio (which we have used to try Deepseek and other models)flame.CPP or VLLM, which allows you to try it at home without connecting to the cloud or depending on OpenAi. Of course, it is recommended to have some technical experience to put everything in motion. It is compatible with tools such as LM Studio, Llama.CPP or VLLM. Where do I download it? Is it really free? Weights are available for free download mainly in Hugging Faceand can also be deployed through platforms such as Azure, AWS or Databricks. The two variants, 120b and 20b, together with the necessary documentation are available. OpenAI publishes GPT-Oss under Apache 2.0 license, which means that you can use it, modify it and adapt it for commercial purposes. Do not pay for download or use it at home. If you choose to execute it in the cloud or on a payment infrastructure, the cost will depend on the supplier you use. Is it easy to use or do you have to have a certain level of knowledge? Using GPT-Oss is not as complicated as training a model from scratch, but it is not a solution that works with two clicks. The models include clear instructions and are integrated into platforms such as LM Studio or Llama.cpp, which allow them to be executed at local ease. That said, you will need to understand basic concepts such as execution environment, models loading and parameter configuration. If you have already worked with other open weight models, it will not cost you. If not, there are guides available, but you will have to devote some time to put it to function as it is. What can this model do? GPT-Oss is a multi-support model. You can reason, generate text, write code, answer questions, search for information and operate external tools, such as python browsers or interpreters. All this is also compatible with agents based on workflows, useful for more complex and automated tasks. Openai states that the 120B model achieves a proximity with O4-MINI in reasoning tests, while the 20B model offers similar results to O3-MINI in Common Benchmarks, with a good performance in programming, reasoning for steps and use of tools. It is not the most powerful model of the house, but one of the most versatile that can be used without paying or depending on your API. Images | OpenAI In Xataka | A watchmaking pump threatens the most powerful alliance of technology: Microsoft’s AGI clause Images | OpenAI In Xataka | A watchmaking pump threatens the most powerful alliance of technology: Microsoft’s AGI clause

Microsoft and Openai AGI clause

by

Microsoft is losing its advantage in AI before its own partner. OpenAI, the company that received 13,750 million dollars of investment, now negotiates from a position of strength to weaken the most lucrative relationship of the technological sector. That of a marriage of convenience in which both parties have more and more misgivings of the other. Why is it important. A clause buried in the contract between the two companies could dynamite the alliance. Yes OpenAi declares having reached the General Artificial Intelligence (AGI)Microsoft would lose access to all future models. It would be anchored in obsolete technology while its partner would conquer the market with higher tools. The panoramic. What began as a marriage of convenience has become a battle for control: Microsoft needs Openai technology to compete with Google and Amazon. OpenAI needs Microsoft servers to train their models. But Each company now seeks to reduce its mutual dependence While renegotia a ratio valued in hundreds of thousands of millions. In detail. The AGI clause works as a three -part watchmaking pump. First, the Openai Council can unilaterally declare that the AGI has reached. Second, there is an economic threshold: If OpenAI shows that its models can generate more than 100,000 million in benefitsYou can cut access to Microsoft. Third, Microsoft is prohibited from developing AGI on its own until 2030. When the agreement was signed, Microsoft thought the AGI would take decades to arrive. Sam Altman, OpenAi CEO, It has been shortening the deadlines. It sounds as if Altman were moving the goal to call AGI to a simple better model to get rid of Microsoft. What has happened. The problems began in November 2023, when Openai’s Council fired Altman without warning Microsoft. Although they replaced him days later, Microsoft lost confidence in its partner. In March 2024, Microsoft hired Mustafa Suleymanfounder of Deepmind, to lead his division of internal. OpenAI responded by diversifying its suppliers Cloud With agreements with Oracle and Google. The figures. Openai folded its annual income, from 5.5 billion to 10,000 million dollars, in 2024. This year It already goes for 12,000 million. But the company continues to lose money: in 2024 it registered losses of almost 5,000 million. Microsoft, meanwhile, generates 75,000 million annually with Azure, and an important part comes from OpenAi. The context. OpenAI wants to restructure your business model to eliminate the top to the benefits and allow investors and employees to have direct shares. Microsoft must approve this change, which gives negotiation power to review the entire relationship. Between bambalins. Tensions have climbed beyond financial aspects. Some OpenAi senior researchers resist delivering their developments to Microsoft, despite the company’s contractual rights until 2030. OpenAI has begun to offer its services directly to business clients, short -circuiting Microsoft as an intermediary. Yes, but. Both companies remain deeply interconnected. Microsoft has built its entire AI ecosystem (Co -cilot, Azure Openai Service) around OpenAi technology. Chatgpt has 900 million downloads to The 100 million co -pilotturning Openai into the real winner of the consumer market. By win. And now what. Negotiations advance towards an agreement that could be completed by the end of summer. Microsoft seeks to keep access to OpenAi technology even after AGI and obtain a 30-35% participation in the restructured company. OpenAi wants freedom to choose suppliers Cloud and offer services through AWS and Google Cloud. Turning point. If they do not get an agreement, Microsoft could keep the current contract until 2030, but would run the risk of losing access to the most important advances in AI. OpenAI, meanwhile, needs Microsoft’s approval to complete its restructuring and access 40,000 million in new financing. The big question It is if one of the most powerful alliances of technology can survive when both partners have become rivals. The answer will also serve who will control much of the future of AI. Outstanding image | Xataka In Xataka | All AI companies promise that the AGI will arrive very soon. The problem is that chatgpt is not the way

GPT-5 will be better than its predecessors. The point is that for Openai that is no longer enough

by

It smells in the environment. GPT-5 is falling. Well, not only is he smells: it is also Sam Altman already shown that He is using it. There are many indications that point to an imminent launch, and it is therefore good time to assess what we expect from this new foundational model and its impact for the industry. Downward expectations. In general, the sensation of media and analysts coincides: the quality jump that we will see from GPT-4 to GPT-5 will not be much less that we saw between GPT-3 and GPT-4 when the latter He threw himself In March 2023. The reason is simple: climbing (dedicate more computation capacity and more training data) It no longer raises such notable improvements like those raised. Expected deceleration. That “Founder of AI“It was already expected – engineers knew that this better rhythm was not sustainable – and precisely the rescue came things like modes of reasoning. To make the model “reason” work: That allowed not to respond immediately, and instead the model explores several paths and ends up deciding the best. The problem? If “reasons” it takes longer to answer, and if it takes longer to answer conversing with it it is no longer useful. And that’s where GPT-5 enters. The off -road model. If we expect something from GPT-5, it is something like “the model to dominate them all.” So far there was a real chaos in OpenAi and its models: we have chatgpt as a unifying layer, but from this chatbot you can choose different models and different ways (reasoning, research, etc). GPT-5 is expected to automatically adapt to user’s needs (although being able to select modes manually does not seem a bad idea for advanced users) and becomes an off-road model that Unify everything. Better in everything, but maybe not too much. Of course, GPT-5 is expected to represent a sensitive leap in the capacities of this founding model. There are many rumors about what will be especially trained to program and to solve mathematical and even scientific problems. Advanced reasoning. It is also expected to have an advanced reasoning capacity, and here there may be two psible striking improvements. The first, that the model reasons better but also does it faster and more efficiently. Or that reason better (and slower) but reason as well as now, but at the same time that does not prevent chatting with it in that way. In both cases the result will be the same: more precise and detailed responses. Deep Think. The second, which perhaps Openai implements the same idea that Google has already implemented in Gemini: a Model “Deep Think” that combines the power of several instances of the same model so that they all reason and end up reaching the best conclusion. This second option, of course, raises the arrival of an even more expensive subscription plan, because this mode of reasoning is very expensive and using it intensively will be oriented only to very specific scenarios in areas such as the scientist or academic. Agentic. OpenAi already recently presented Your chatgpt agentand GPT-5 is expected to have a key participation in the future of that option. Right now, the agthic capacity of that service is surprising – it takes the control of the browser and you ask you to do something – but also limited (it takes a long time in everything, it is often wrong). With the new foundational model its performance should end up improving in a sensitive way too. A model of AI that threatens more jobs. THE IMPACT IN THE INDUSTRY It is certainly uncertain, but if these capabilities are really notable, there will be more threatened jobs. GPT-5 could, for example, being a strong candidate to even replace people in customer service and technical support: their ability to talk in real time-and to remember more and more details of users-is very striking in this regard. Sectors such as health, with improvement of diagnostic prelminar, or education adapting more and better to each student, such as They have already profiled with their mode of studyThey are clear candidates to be affected. After the failure of GPT-4.5, OpenAi needs a total success. A few months ago OpenAi suffered The greatest failure in its history with GPT-4.5a model that cost a lot to launch and did not show really striking improvements. He ended up going back and now he needs GPT-5 to represent a decent (at least) jump. Keep burning money. The role of GPT-5 is also key to maintaining the good reputation of the company. They live in much of it, and demonstrate the gigantic investment rounds that it has achieved over the last years. In Openai they continue to burn money as if there were no tomorrow, and although it already generates respectable income – it is estimated that will enter 12,000 million dollars In 2025— he is far from profitable. Fierce competition. The threat is also multiplied against rivals like Google and now also a goal. The first is Your most direct competitor (together with Anthropic) in terms of quality of commercial foundational models. The second has become a real toothache, especially since it has so much money (and generate both thanks to Instagram and Facebook) that does not need external investment and can waste money without apparent problems. For example, stealing talent without stopping. Please subscribe to chatgpt. Chatgpt’s popularity is undeniable, and today many users associate “AI” with “Chatgpt” when using any of the available models. That has allowed its number of weekly unique users either 700 million “Four times the one who was a year ago,” exceptional growth. Subscriptions, yes, are slower: at this time OpenAI has five million business business users, while the figure was three million in June. He needs many more to deal with his astronomical expenses, and GPT-5 should listen to get it. In Xataka | The AI war intensifies: Anthropic cuts Claude’s access to OpenAi. He has done it before the launch of GPT-5

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.