DeepSeek

Deepseek put China on the AI ​​map. The danger is that this revolution stays in a day flower

by

Deepseek R1 was eating the world At the beginning of the year. This Chinese model, apparently out of nowhere, caused A true shock In the AI ​​industry, but since then there has been movement. Actually there has been one, but the disturbing thing is precisely what that movement has been. Hi, Deepseek v3.1. The startup advertisement Last week the launch of Deepseek V3.1, a new version that stood out for being an improved hybrid of Deepseek V3 (fast response) and Deepseek R1 (reasoning). There was also good news in terms of their performance: according to the Benchmarks published by those responsible, it was significantly higher than their predecessors. Visible (but non -dramatic) improvements. In the “model card” (model card) that those responsible offers In Hugging FaceDeepseek v3.1 (in reasoning mode) proved to behave slightly better than Deepseek R1-0528, —Your previous version, more powerful-in areas such as programming or in mathematical tests, but some users who have tried it there comment That except in those areas, the model is worse and “it behaves poorly when following instructions or prompts provided by users.” Others confirm it and They assure which is useful for programmers, but not for other areas. It also has limitations on its multimodal support, and focuses on the text instead of providing more options for another type of interaction, for example from voice, image, video or audio messages. A Chinese model for Chinese chips. But even more interesting it was that Deepseek V3.1 has been designed and launched with a clear objective: avoiding the dependence of foreign chips. The FP8 precision used makes this model behave very well In the next -generation Chinese chips. The strategy seems very interesting for the startup, which could thus have a very aligned model with the priorities of the Chinese government. This is: use local models for local chips as much as possible. And R1, what? From there some doubts arise. The first, which affects Deepseek R1, the model with which the startup “broke” the market at the beginning of the year. The company has eliminated all references to this model in the characteristic of “deep thought”, which has generated doubts about the potential appearance of its expected successor, a hypothetical Deepseek R2. Loses users. But while that theoretical model comes – if it does – the company faces a more immediate threat. As they point out In SCMPDeepseek is losing users (or at least relevance) in recent months. In the first quarter of the year its market share within the scope of the IA Open Source models used on the PPIO cloud platform was a spectacular 99%. However, in the second quarter that percentage has dropped to 80%. Fierce competition. That fall relevance has an obvious reason: its local competitors are squeezing. And a lot. Among them is the family of models Qwen from Alibaba, but Also others like Kimi-K2-Instructof the startup Mosohot AI – in which Alibaba has also invested – which is becoming one of the most popular models of recent weeks. Delays and deceleration. Precisely the focus on being able to make the most of future Chinese chips seems to be the reason that this hypothetical Deepseek R2 is being delayed. At least that is the hypothesis that consider In Financial Timeswhere they revealed that the startup has failed when trying to train with Huawei chips. The situation has made them Training with Nvidia chipsand that are using the Huawei Asce for the inference stage, that is, the interaction with the model via web or API by users. But this attitude is “very Chinese”. We may in Western countries we are accustomed to a much more frantic pace and that we expect constant updates and improvements with an eye on the short term. In China, philosophy is usually the opposite, and companies adopt A long -term strategy even if immediate benefits are lost. Maintaining a low profile is also usual among those companies, which try not to make much noise … until they do, as Depseek has already demonstrated. Thus, we will have to remain very attentive to the activity of this startup, because surely he will be working to continue being one of the protagonists of the AI ​​panorama. Image | Tim Reckmann In Xataka | Deepseek has suggested that Nvidia chips no longer needs. We believe to know who is buying them

Deepseek has suggested that Nvidia chips no longer needs. We believe to know who is buying them

by

Deepseek shook at the beginning of 2025 the foundations of the industry of the artificial intelligence (AI). This Chinese model developed by the specialized quantitative coverage fund trading High-Flyer algorithmic monopolized all the attention because He gave us free of charge an AI with a quality similar to the comparable solutions of Openai or Google. However, this was not all. And it is that for three weeks the debate about the hardware used by this company to train its model. High-Flyer, who specializes in addressing investment decisions using advanced mathematical models and computational algorithms, says he trained Deepseek R1 using 2,048 chips H800 of Nvidia. Several analysts They soon react ensuring that he had actually used 50,000 GPU H100which are more powerful, bought through intermediaries. The sanctions of the US government prevent Nvidia According to SCMPDeepseek’s next iteration will dispense with the GPUs of this American company. High-Flyer has suggested that everything you need already finds it in China Last week those responsible for the Deepseek development published on the Wechat platform an entry in which they hinted that next generation GPUs for China They will launch soon. This comment has unleashed endless speculation about the name of the Chinese company that is going to make this announcement, but, above all, it has put on the table the possibility that High-Flyer is preparing the following Deepseek iteration using these chips. A priori does not seem at all crazy. According to SCMP The five GPU designers for the Chinese who have the ability to deliver to High-Flyer the hardware they need are Huawei Technologies, Cambricon Technologies, Moore Threads, Hygon Information Technology and Metax Integrated Circi. Any of them could be responsible for the announcement anticipated by Deepseek developers in Wechat, but we We bet on the first three Because, in our opinion, they are Those who are “in a better way”. Cambricon Technologies is one of the companies specialized in the design of GPU for AI with greater growth potential Although it is not as well known as Huawei or Moore Threads, Cambricon Technologies is one of the companies specialized in the design of GPU for AI with greater growth potential. In fact, he has received the approval of the Shanghai bag (China) to raise 560 million dollars. Will allocate them to the design of four chips for training and inference of AI models, and also to the development of an alternative to CUDAfrom Nvidia. On the other hand, Moore Threads He has developed several GPU for AI applications that, on paper, rivaize some of the advanced solutions that have placed in the Nvidia, AMD or Huawei market. MTT S4000 and MTT S3000 cards They are its most interesting proposals right now, although, curiously, in its porpholio the MTT S80 card also appears, a proposal for games and content creation that, according to Moore Threads itself, has a calculation capacity of 14.4 Tflops in single -precision floating coma operations. The other indispensable actor in the Chinese chips industry for IA is Huawei. His most ambitious proposal right now is the chip Ascend 910dwho seeks to overcome the performance of the GPU NVIDIA H100. However, this Chinese company has also recently presented its chip Ascend 920a solution that is clearly destined to occupy in the Chinese market the gaps that the NVIDIA H20 GPU is going to leave. This proposal will enter large -scale production during the second half of 2025 using 6 NM integration technology that have presumably developed elbow with Huawei elbow and SMIC (Semiconductor manufacturing international corp). More information | SCMP In Xataka | Nvidia has to deal with the absolute distrust of several US legislators. His plan in China is in danger In Xataka | The US wants to end the chips for the Chinese that are sold abroad. And China knows how to defend oneself

Deepseek marked a turning point in the AI race. Now another Chinese company wants to imitate its success: Kimi K2 is born

by

The Chinese startup Monshot AI has presented Kimi K2, an open -source artificial intelligence model that arrives with outstanding programming capabilities and autonomous tasks that, according to The published benchmarksThey spray competition in several of their models. Its launch occurs at a key moment for the sector, when Chinese companies seek to replicate the disruptive success of Deepseek with potential height models and much cheaper than market alternatives. Kimi does not come from nothing. MoNshot ai was one of the most promising startups in the Chinese ecosystem of AI and that giants like Alibaba have invested greatly. His Kimi chatbot reached third place in monthly active users in August 2024, but fell to the seventh in June After the emergence of Deepseek R1 in January. Now try to recover ground with a strategy that combines open source and aggressive prices, following the formula that catapulted Deepseek. Image: MoNshot AI What Kimi K2 offers. The model has 1 billion total parameters and 32,000 million activated parameters, using The well-known Mixture-Of-Experts architecture to optimize computational costs. It is presented in two versions: a base for researchers and developers, and another optimized for conversation and autonomous tasks. Kimi K2 thus becomes Moonshot AI’s proposal with the ability to act as an intelligent agent to use tools, write code, complete workflows or talk, among other tasks. Kimi K2 explained in numbers. In performance testsKimi K2 has achieved 65.8% precision at Swe-Bench Verified, one of the most demanding benchmarks for software engineering. In LivecodeBench it reached 53.7%, exceeding 46.9% of Deepseek-V3 and 44.7% of GPT-4.1. In mathematics, its 97.4% score in Math-500 exceeds 92.4% of GPT-4.1, suggesting significant advances in mathematical reasoning. The price factor. MoNshot is charging $ 0.15 per million input tokens and $ 2.50 per million tokens out of the developers who use their API. Compared, Claude Opus 4 It charges 100 times more for the entrance (15 dollars) and 30 times more for the output ($ 75), while GPT-4.1 charges 2 dollars per entrance and 8 per exit. In addition, the model is available for free in Web applications and Kimi mobile, without monthly subscriptions that require chatgpt or Claude for their most advanced models. Technical innovation. MoNshot has developed the MuCanclip optimizer, which allows train models of one billion parameters “With zero training instability.” This technology could drastically reduce the training costs of large models, a problem that has limited the development of AI to companies with greater resources. Double channel strategy. The company offers so much Free access to the source code as payment API at a very competitive price. This strategy allows companies to start with the API for immediate implementation and then migrate to self -healing versions either by regulatory cost or compliance. And it is that each developer who downloads Kimi K2 becomes a potential business client. Moment of inflection. Kimi K2 represents a convergence point where open source models and proprietary alternatives shake hands. MoNshot AI intends to turn Kimi into a tool for everything, while offering its open source model and is reserved to charge for the use of its API for all types of implementations. And now what. The launch reaches a critical point in which both Openai, such as Google or Anthropic, must respond to this wave of cheap and high quality language models. The issue is no longer whether open source models can match the owners, but if large technological ones can adapt their business models fast enough to compete in this new scenario. The looks are put in GPT-5 And in the next movements of the industry at a rate, as always, accelerated. Cover image | Xataka with Mockuuuups Studio and Kimi AI In Xataka | Grok 4 destroys the tests and aims to be the most advanced AI model. The problem is that Elon Musk continues to sabotage his answers

After the emergence of Deepseek, the “seven magnificent” of the Tech industry have collapsed in the stock market. All except Apple

by

The year began well for Nvidia. On January 29, 2025 its capitalization I reached The 3.49 billion dollars and everything seemed to go on wheels. The Surprise arrival of Deepseek R1 It changed things a lot and joined other factors to cause spectacular collapse. Two months later, this Nvidia market capitalization is 2.77 billion dollars: it is almost 21% less. That effect has been contagious, but one of the greats is falling the storm. Apple. As they point out In five daysthat January 25, 2025 Apple had a capitalization of 3.55 billion dollars, and at this time that value is 3.35 billion, 5.6% fall. Sensitive, of course, but much less than that of its rivals of the group of “The Magnificent Seven”. That they have stayed in … Not so magnificent. Next to the fall of Nvidia are those of Microsoft, Alphabet, Amazon, Meta and Tesla – as we say, is saved a little. If we analyze the evolution of market capitalization of the seven the performance of these two last months, the “average” drop is 13.5%. They have lost more than two billion dollars compared to 15.58 billion dollars in late January, a real collapse. It’s not just care. The impact of Deepseek has not been the only factor that has contributed to those falls. They have had a lot to do The recent tariffs That is imposing Trump to imports of all kinds of products – foreign cars They are the last victims-. These taxes and Trump’s protectionist policy are forcing many companies to restructure their strategy, and investors – and consumers – are clear what the impact of all this will be: price increases everywhere. Why does Apple endure? Of the great technology, Apple is the only one that has managed to mitigate the losses relatively. Probably partly because of his “warm” attitude to AI. Your interest in data centers fever It is practically nulland despite the Recent criticism It is clear that it is not “burning money” as other companies in the sector do. The rest of the group has invested true fortunes In this segment, although some They are stopping. Bubble in sight? These days are 25 years of the bubble of the Puntocom, and what is happening with the great technology and the AI ​​segment does fear for an AI bubble. There are certainly similarities between both situations, but also important differences. Apple, especially solid. Cupertino’s company is usually More immune that their rivals to these fluctuations in the world of finance. In the face of complaints about the relative lack of innovation or New disruptionsApple has managed to diversify income – especially with the expansion of its services – and continues to maintain confidence of both investors and users. Image | Zhang Kaiyv In Xataka | Deepseek R1 is not just another AI model: it is the greatest existential threat that Silicon Valley has faced

Huawei has just said goodbye to Android by Deepseek. And this strange folding is the first step

by

I had announcing the movement For a few months, and it has finally materialized on a phone. Huawei He just said goodbye to Android in his native countryand he has done it with a curious phone with an never seen aspect relationship. This is the first phone that lands on the market with Harmonyos Nextthe operating system completely developed by Huawei with its own kernel and without a trace of Android. In addition to premiere operating system, it is the first Huawei device to release Harmony Intelligence, its AI in collaboration with Deepseek. Goodbye to Android. The day came. Huawei has presented its first phone without a trace of Android, the Huawei pure x flip. To date, the company’s mobiles were marketed in China with Harmonyos. An operating system that, although the company wanted to completely disconnect from Android, was based on its nucleus. With Harmonyos Next things are different. This is an operating system developed from scratch, does not use Android code, is not compatible with its apps (at least, not if these are not modified) and uses its own core (HMS or Huawei Mobile Services) instead of the Google Application Service. This is the first 100% Harmony mobile. The first step for AI. Harmony Intelligence It is your alternative to Apple Intelligence and Samsung Galaxy AIpromising to go beyond an assistant for AI. This AI has trained both the Pangu model, developed by Huawei, and in Depseek, a model that is permeating quickly in Chinese mobiles. “We have trained our AI assistant for millions of hours with our large -scale language models and we have taken that intelligence to the device for the first time, so that it looks like a human and interact naturally, with conversations with different emotions and detecting the user’s mood.” As Richard Yu, CEO of Huawei promises, the focus is that the person-one interaction is as natural as possible, coming to promise that the assistant will be able to detect user’s emotions (it has not been detailed how). In a more landed plane, it will be able to eliminate photographs objects, generate and modify them, convert voice, and even describe the objects that the telephone camera sees, so that people with accessibility problems can “see” through the mobile. Another especially curious function is to be able to read books without touching the phone: we can pass pages only with our eyes. Huawei wants a 100% own ecosystem. The key here is not that a Huawei mobile releases its AI, is that Huawei is already prepared, in agreement with Deepseekto compete with the greats of the industry. Harmonyos Next is a platform oriented to all types of devices, from home to the connected car, and this first jump presenting Harmony Intelligence is key to the Chinese manufacturer. The company has advanced that its first platform for PC (Harmonyos Next brought to desk) will land in May, and it is expected that throughout the year we will start to see this operating system in virtually all new products that launch the market. The pure huawei. The lucky one who has released Harmonyos Next is the pure x, a folding phone in appearance ratio 16.10 when it is deployed. It is a telephone that, being folded, has a 3.5 -inch screen with 980 x 980 resolution, 2,500 peak shine nits, LTPO technology and appearance ratio 1.1: completely square. Upon opening, a curious 6.3 -inch panel is displayed with resolution 2,120 x 1,320, a completely different format than we are used to. The camera system is triple, with 50 MP for the main camera, 40 MP for ultra wide angle and 8 MP for 3.5X teleobjective. It has a 4.720MAH battery with a fast charging of 60W and will be sold in China at a price that starts from 7,499 yuan (just over $ 1,000) in its 12 + 256 GB version. At the moment, far from Europe. If the question is whether Huawei’s plan to completely end its Android dependence, the answer is no … for now. In Europe, the company is looking for alliances with developers after projects such as Aurora Store to facilitate the installation of Google services in its terminals (something that is easier than ever), and betting on EMUI based on ancient Android versions. The key to Harmonyos Next ends up being global is in the Chinese manufacturer alliance with the rest of the developers and technological. In China, Google apps or giants as a goal (Instagram, WhatsApp and Facebook) have hardly any presence, in our territory they are an essential for a good part of the users. Currently, they can be installed in EMUI thanks to the microg Google Services framework. Without it, the matter is complicated. Image | Huawei In Xataka | China has a possibility to unseat Android forever. One called Harmonyos

While OpenAi insists on being the new Google, Deepseek says they have higher goals: AGI

by

Sam Altman has been very clear about his way of seeing things in Openai from the beginning. He has always been in a hurry to launch the ED models his company, market them and be looking for investment rounds to be able to sustain that frantic expense rhythm. The funny thing is that one of his greatest rivals, Deepseekyou are looking for a totally opposite strategy. Income grows. As they point out In Financial Timesthe company led by Liang Wenfeng has become one of the favorites of the Chinese technological panorama. Its payment services, consisting of the API for the use of Deepseek V3 and R1, have worked very well and have allowed income to cover operating expenses for the first time last month. But Deepseek doesn’t look for money. However, industry experts point out in FT that Liang has no intention of trying to take advantage of the time to maximize financial income marketing more and more their (efficient) products. No investment. Nor seek investors who support their goalsand in fact there is talk of how difficult it is to talk to the founder of the company. An investor of a multi -million dollar fund in China indicated how “we took advantage of high -level government connections and we only managed to sit with someone from their financial department who told us that they regretted it but did not look for investment rounds.” A for the AGI. Instead the company is focused on developing its current models and also in seeking the development of general artificial intelligence (AGI). It is the same challenge that other rivals pursue, but it is not clear what path each one is following to reach those with superior abilities even to those of human intelligence in all areas. Totally different from OpenAI. Deepseek’s approach is totally opposite to Openai, which from the first moment took advantage of Chatgpt to create a commercial business around that chatbot. Then the successive (and colossal) rounds of investment would arrive. The rumored SoftBank’s round, which It is estimated at 40,000 million dollarswould make its value amount to 260,000 million dollars. A small draft startup. In Financial Times they also point out that Depseek has 160 employees, something that contrasts with the more than 2,000 that OpenAi has. That has made others like Alibaba or Tencent convince business clients such as Apple, who will use QWen (from Alibaba) as an option for iPhone users in China. And with a “small” infrastructure. According to sources close to the company, Liang bought 10,000 GPUS NVIDIA H800 and 10,000 A100 in recent years. He did it before access to them was vetoed. It is not an too high amount if we compare it for example with which Elon Musk acquired In XAI, and also both are somewhat less powerful models than H100 and of course than modern B200. Deepseek R2 and V4 are on their way. The new versions of their AI models are in full development. They were expected to be launched in May, but the company may decide to accelerate that launch and make them available to the general public even before to take advantage of the good time of the company. Outstanding image | Solen Feyissa In Xataka | Deepseek, in the spotlight of European regulators: Italy and Ireland act against privacy concerns

Deepseek closes external investments and risk capital. You have three reasons to do it

by

Deepseek, The great sensation of the beginning of the year in AIhe is breaking with all the rules of the startup ecosystem not only for his efficient AI modelbut also for rejecting the risk capital that other competitors need such as eating. Its founder, Liang Wenfeng, maintains 84% ​​of the property (an anomaly in its field) and does not seem to be in a hurry to give control. Chinese technology broke into the industry in January With its new AI model, but unlike its competitors, it does not announcing multimillionaire financing rounds. Nor want to do it, according to anti The three reasons. Liang Wenfeng has clear reasons to keep investors at bay: 1. He does not want to lose control of his long -term vision on AI.2. Depseek has sufficient its own financing through its High-Flyer investment fund.3. He fears that external investors, especially Chinese, intensify concerns about privacy and security. Why is it important. In a sector where competitors are launched to capture thousands of millions to finance the expensive AI race, Depseek is betting on an alternative path. Financial independence also gives Deepseek freedom to focus on research and development, instead of seeking rapid monetization that investors usually demand. Between the lines. Liang does not usually hide his distrust of investors. In a 2023 interview He openly criticized the obsession of venture capital funds by rapidly monetizing AI, to the detriment of advanced research. This position symbolizes a growing skepticism in the technological sector about whether the traditional financing model is compatible with the development of long -term transforming technologies. The figures. The Deepseek owned structure is unusual for such a powerful startup: 84% is owned by Liang Wenfeng, the founder. 16% are in the hands of people linked to their investment fund, high-flyer. 0% is of traditional external investors. In detail. Liang has funded Deepseek with the benefits of High-Flyer, its quantitative investment fund. “Money has never been a problem for us; the problem is in the prohibitions for sending advanced chips,” he said in 2023. This financial self -sufficiency has allowed Deepseek to develop without the usual external pressure of investors, usually focused on short -term growth metrics. The backdrop. As a Chinese company, Depseek operates under laws that give the Government broad access to their data. This has already caused prohibitions of use in several countries and private companies. The entry of Chinese investors could get this situation worse. The US government has history by sanctioning Chinese technology with government ties, such as Huawei and DJI. And now what. Deepseek will need more and better AI chips to stay competitive, according to Liang himself. These components are expensive and are strongly restricted in China due to the export controls of the United States. Without external financing, Depseek could be left behind in the technological career against rivals with more resources, such as OpenAi or Anthropic. Of course If someone has demonstrated ingenuityIt is Deepseek. In Xataka | Manus is the new sensation of China after Deepseek. Is generating as many expectations as doubts Outstanding image | Deepseek + Philipp Katzenberger | Alejandro Luengo

Manus is the new sensation of China after Deepseek. Is generating as many expectations as doubts

by

Last week voices began to hear in the AI ​​segment. We theoretically had a new Deepseek phenomenon. It was Manusa theoretical AI agent who raised especially striking options. However, in social networks there have been doubts and suspicions about the real capacity of this model. The unknown about its real value is there. Manus is not an AI agent. Those responsible on its official website is “a general agent who connects ideas with actions: not only thinks, delivers results.” It is actually more a “deep research” mode – like the ‘Deep Research’ of OpenAi – about the data that we give. Thus, Manus is theoretically capable of creating you a detailed 7 -day travel plan by Japanhelp you perform a deep analysis of the Value of Tesla’s actionscreate for you a presentation for a class on the conservation law of the momentFor example. Expectations in style. One of the product managers in Hugging Face He described it as “the most impressive tool I have ever tried.” His Official Discord server He has attracted almost 170,000 users since his creation a week ago. The expectation is remarkable, but as they explain In TechCrunchthere are doubts about their true capacity. Manus is based on Claude and Qwen. There are users who They have already indicated that manus seems not to have been created from zero, and apparently it is Totally based in varied models such as Claude 3.7 and Qwen (Alibaba). One of his co -founders, Yichao “Peak” Yi, confirmed that are in effect based on other products and they are a layer that relies on these models. In the presentation video Yichao stressed that In some benchmarks Manus was superior to Deep Research already OperatorOpenai products. It is? Errors and inconsistencies. Some experts such as Alexander Doria, co -founder of the startup of the Pleias, have indicated that manus makes mistakes and infinite loops When using it. Others point out how Make mistakes With questions with clear answers and Do not cite sources consistently. Beta by invitation. For now it is not possible to access manus options, and it is only possible to use it if any other user invites you. In TechCrunch they have managed to have access, and the tests performed, explained the editor who tested it, “have not been especially positive.” None of the agerentic tasks that he tested – seek table in a restaurant, take out plane tickets – worked. In The Register they have also had access and Nothing have been impressed with the results. “They tried to create a Mario clone that turned out to be quite terribleFor example. Other users who have tried or They praise it or They believe that despite its lights and shadows is a special promising product. At the moment this is not a Deepseek. The truth is that the rapid popularity of Manus contrasts with that of Deepseek, which was available from the first moment as an Open Source development that anyone could download and use. The company may want to take advantage of that expectation generated by betas with invitation access and prefer to avoid greater infrastructure expenses. In Xataka | China has a plan to achieve scientific supremacy in full struggle with the US to lead in AI

Chatgpt costs $ 700,000 a day, Deepseek just 83,000. A key is huawei and play cards differently

by

2025 began with a tsunami in the segment of the artificial intelligence. After months and months talking about different models and companies such as Google, Microsoft, Apple, Goal And, of course, OpenAIa Chinese company took off the manga Deekseekan AI that shook industry foundations. Beyond its possibilities or how good it worked, which stirred the waters were economic and hardware issues. Almost from the beginning the question of How China had taken out an AI like Deepseek With the hardware limitation they have due to the Commercial War with the United States and the impossibility of buying the most powerful graphs of Nvidia (although with controversy). To do this, the company defended that it had to pull ingenuity Thanks to an infrastructure of NVIDIA H800 chips and a training of more than 2,788 million hours with a ridiculous cost: 5.6 million dollars. And it seems scarce because Openai He invested about 100 million dollars to train GPT-4. Another melon is what it costs to keep it. As noted ReutersIf Chatgpt costs about $ 700,000 a day, Deepseek drops to $ 87,000. And here some things that have to be taken into account. Deepseek is 10 times cheaper to maintain than chatgpt, according to Depseek Last Saturday, and as noted ReutersDeepseek revealed some data on costs and income related to its V3 and R1 models. The first is a traditional, more conversational chatbot, resulting ideal for the writing and creation of content. R1, however, is a reasoning model. It stands out solving problems, using logic and is able to show step by step reasoning, using continuous learning. To compare with better known models, Depseek V3 would be like GPT-4 and R1 something similar to OPENAI O1. In the report, Reuters highlights that the theoretical relationship of Deepseek-benefit costs is up to 545% per day. Of course, the company itself warns that real income is significantly lower, but another pearl they have left since Deepseek is the cost of maintenance. Keep chatgpt working costs about $ 700,000 a day to OpenAI (at least two years ago). The reason is that the infrastructure of Microsoft Azure Serversit has a considerable energy cost, you have to pay wages and, obviously, all the power in hardware to process the consultations you receive every second. In yellow, the costs. In blue, theoretical income Deepseek “only” 87,072, a ridiculous price compared. A few days ago affirmed That, renting the H800 costs less than two dollars per hour and the estimated theoretical income is just over 560,000 dollars, which would add more than 200 million dollars in a year. In the upper graph, Depseek shows the cost of maintaining R1 and theoretical income thanks to the tokens that are generated, whose price depends on the moment of the day, being cheaper at night. They also clarify that Deepseek V3 is “significantly cheaper.” This opens more issues. One is how it is so cheap because training an AI, of course, is not. Ignoring the accusation of theft by OpenAIif Deepseek has not deflated the numbers, puts on the table a situation in which it is not needed so many graphic power to train an artificial intelligence. Here the key is the ‘reinforcement learning’, the way Deepseek has found do with much lessbut it should also be noted that, although during training the Nvidia chips are used for the R1 model, in the inference it is using The Ascend 910b of Huawei. Huawei chips are cheaper And, supposedly, more efficient, and this decision of Deepseek is almost more relevant than what may cost to maintain the system. The reason is that you can teach the rest of artificial intelligence companies that, perhaps, It is not worth using the latest generation GPUs For everything, but only for training that occurs sometimes counted, before the implementation of AI, and then use other more efficient and cheap GPUs for inference. This inference is what is done later, in a use that we could call “real” The training would be the equivalent of swallowing technical manuals in a five -year career and the inference such as implementing that knowledge and reasoning starting from the base you have and without having to learn them again. In the end, the controversy of five million dollars Deepseek is going to be there for a while, especially when we compare with OpenAi numbers, but it is clear that Depseek is doing things from another approach and can be a good mirror for companies that come behind. And, with one China very focused on the development of both AI as of Hardware for AIit can be the perfect ‘spearhead’ model. Images | Github (Deepseek), Xataka In Xataka | Deepseek has created another Milmillonaria fortune: Liang Wenfeng has become popular but its wealth is still a mystery

The Singapore government has revealed which companies are involved in the delivery of GPU from Nvidia A Deepseek

by

Depseek continues to be the artificial intelligence (AI) of the moment five weeks after its irruption. And is that the debate about the hardware used by this Chinese company to train your AI model Keep on the table. High-flyer, your parent company, is A quantitative coverage fund specialized in trading algorithmic. This simply means that this company uses advanced mathematical models and computational algorithms to address investment decisions with the greatest possible success guarantees. Deepseek was born as a high-flyer secondary project to take advantage of its computer resources and “put one foot” in the AI ​​industry. Its creators say that in the training of their model they have used only 2,048 chips H800 of Nvidia. However, Some analysts defend that, in reality, its infrastructure brings together 50,000 GPU H100 bought through intermediaries. This is the problem. High-flyer could legally buy the H800 chips to the entry into force of US sanctions of November 16, 2023, but the H100 GPUs should not be in their possession. Singapore is the entrance door to China of the chips for the most advanced Nvidia The US government has suspected for many months that Chinese companies and research centers dedicated to AI acquire the most advanced NVIDIA GPUs through Singapore and Malaysian intermediary companies. This possibility is no longer just a hypothesis. And it is that the Singapore government has confirmed that it has identified those responsible for diverting to China, and presumably towards the Deepseek parent company, servers that contain the high -performance GPUs produced by NVIDIA. The US now has the opportunity to tighten its fence a little more about China This information was revealed last week by the television channel Channel News Asia, and today the Minister of Internal Affairs and Justice of Singapore, K. Shanmugam, He has confirmed it. Interestingly, it has not specified what the GPUs that incorporate these machines, but it has made public a very important fact: The name of the companies They have manufactured the servers. And they are two very important Nvidia customers: Dell Technologies and Super Micro Computer. If it is finally confirmed that Depseek, or any other Chinese company that is dedicated to AI, is getting the Nvidia avant -garde GPUs acquiring servers of these companies in Singapore or Malaysia, USA will have the opportunity to tighten your fence a little more. However, this would not demonstrate the guilt of Dell and Super Micro, although its indirect involvement in the traffic of the Nvidia chips. This circumstance would put on the table the need to control with more precision where their servers will stop, something that, on the other hand, It is not easy. Whatever the Singapore government claims to be willing to collaborate with his American counterpart to end the illegal vanguard GPU traffic. Image | Nvidia More information | Reuters In Xataka | We can forget an AI without hallucinations for now. The general director of Nvidia explains why

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.