competes

competes with all of Asia. And he is losing

by

The race to develop models of artificial intelligence (AI) that the US and China maintain in their struggle for world supremacy is entering a new phase. And the physical applications of AI are gradually gaining prominence. By “physical applications” we mean the integration of one or more AI models into a mechanical device, such as a humanoid robot or a machine for industrial automation applications. The interesting thing is that the rules that will determine which countries will lead this emerging industry are different from those that currently govern competition in the field of AI models and software applications. And they are so for a reason: the robustness of the supply chain plays a fundamental role when the product stops being immaterial and takes the form of a cutting-edge hardware solution. In this scenario, Taiwan, Japan, South Korea, and, above all, China, have a great advantage over the US and Europe. China is committed to the strategy with which it has managed to dominate the electric car Before continuing to investigate the industry specialized in the development of physical applications of AI, we are interested in diverting our gaze for a moment to the semiconductor market. At the moment in Asia they are manufactured 90% of memory chips, 75% of microprocessors and 80% of silicon wafers. These figures are overwhelming. So much so, in fact, that in the medium term they remain absolutely out of reach of the West. It will be very difficult for the US and Europe to compete with the Chinese companies UBTech Robotics, Agibot or Unitree Robotics The electric car market also clearly illustrates the strength of Asia in general and China in particular. The country led by Xi Jinping leads it with a global market share of 62%. On the other hand, the South Korean companies LG Energy Solution, SK On and Samsung SDI accumulate a joint global quota in the market for batteries for electric cars of 16%. And the Japanese Toyota produces more than one million hybrid vehicles per year. Japan’s position is peculiar because leads the production of hybrid carsbut it is lagging behind in the pure electric car market. Whatever China’s strength in the electric car market and Asia’s leadership in the integrated circuit manufacturing industry have been built on a very robust supply chain and overwhelming production capacity. Jixun Foo, the senior director at venture capital firm Granite Asia, which specializes in technology investments in Asia, ensures that “AI is not just about models or software applications (…) If we look towards Asia our advantage is in the supply chain, hardware and engineering capabilities.” That’s the key. China has opted for the same strategy in the physical applications of AI industry that has led it to lead the electric car market. Its meticulous control of the entire supply chain, its high manufacturing capacity and its cost optimization in all probability They are giving you a very important advantage in the industry of humanoid robots and machines for industrial automation applications. And not only in its struggle with the United States; It will be very difficult for Europe to compete in a market in which Chinese companies such as UBTech Robotics, Agibot or Unitree Robotics are already fully established. Image | UBTech Robotics More information | SCMP In Xataka | The US is doing everything to drown China. China has already achieved that 35% of its chip machines are its own

DeepSeek has just released a model that competes with Opus 4.6. It costs seven times less and runs on Chinese chips

by

They have passed 484 days since that “DeepSeek moment“, but the wait It seems to have been worth it, because we have the new DeepSeek V4 with us. We are facing an absolutely gigantic open weights model that once again promises to crack the foundations of the proprietary foundational models of Anthropic, OpenAI or Google. This is moving, gentlemen. Gigantic and open. DeepSeek v4 is an Open Source model and comes in two versions. The first is the Pro, with 1.6 trillion parameters (1.6T), of which it has 49,000 million active. The second is Flash, with 248,000 million parameters (248B, huge for a “Flash” model) of which 13,000 are active. More efficient than ever. Both versions they make use of a Mixture-of-Experts (MoE) architecture, which means that only a fraction of the parameters are activated in each inference. This allows the computational cost to be reduced significantly. Both versions support a context window of one million tokens—to include novels and novels at once as input—when in v3 it was 128,000 tokens. Furthermore, this model is much more efficient than its predecessor in computing per token: it requires only 27% of the operations per token and 10% of the KV cache compared to DeepSeek v3.2. Benchmarks promise. DeepSeek’s internal testing reveals that v4 Pro-Max (the best model with the highest reasoning ability) outperforms or is on par with Claude Opus 4.6 Max, GPT-5.4 xHigh, Gemini 3.1 Pro High, Kimi K2.6 and GLM 5.1. The results, however, are not independently verified, which means we should take them with caution. The numbers are still striking: in LiveCodeBench, a programming test, DeepSeek v4-Pro-Max achieves a 93.5% score compared to 88.8 for Opus 4.6 and 91.7% for Gemini 3.1 Pro. In other tests there is more variability, but at least on paper DeepSeek v4 Pro seems as good as Opus 4.7, which until now was the absolute benchmark. Much cheaper. But as happened with its previous version, the difference in price with those models from US companies is astonishing. As point the analyst Simon Willinson, the official prices of DeepSeek v4 Pro are 1.74 dollars per million input tokens and 3.48 dollars per million output tokens, up to almost seven times less than those of Opus 4.7 and up to almost 9 times less than those of the new GPT-5.5. With DeepSeek v4 Flash the cost is 0.14/0.28 dollars per million input/output tokens, when GPT-5.4 Mini costs up to 16 times more. The conclusion is obvious: if it really does what it says it does, the price is an absolute bargain. That is precisely the challenge: that real experience confirms what the benchmarks say. The hardware mystery. DeepSeek has not revealed what hardware has been used to train this version of its founding model. In the past they did admit that they had used NVIDIA’s H800s. Which yes it is known The thing is that the model has been developed to run on both NVIDIA and Huawei Ascend chips. This last has confirmed Baidu that its Ascend Supernode clusters based on the Ascend 950 will fully support DeepSeek v4 versions. Huawei support is “horrible” news for the US. In The Information they already commented that one of the reasons for the “delay” in the appearance of this model was to adapt it so that it worked without problems with Huawei chips. That support is according to Jensen Huang “horrible” news for the US, because it means that dependence on NVIDIA chips no longer exists or at least is reduced to a minimum. But. The launch comes at a difficult time for the company. Guo Daya, one of the people responsible for the v1 and v3 models, has signed for ByteDance to work on AI agents. Luo Fuli, who led the development of v2, joined Xiaomi last year. This launch also coincides with DeepSeek seeking external funding for the first time. They are expected to raise about $300 million and obtain a valuation of about $20 billion. according to The Wall Street Journal. From the surprise effect to the continuity effect. The launch of DeepSeek R1 in January 2025 was surprising because it demonstrated that China could train competitive models at a fraction of the cost of Western models. With DeepSeek v4 that surprise effect disappears to give way to the continuity effect. This model seems to maintain precisely what made the previous model famous: extraordinary power at a very low cost. Bad news for Anthropic. Such low prices are terrible news for Anthropic, which in recent weeks has been forced to execute a kind of “reduflation” of their new modelswhich are not more expensive but consume many more tokens. We’ll have to see if DeepSeek v4 Pro is as good as the company promises, but if it is, we’ll have another “DeepSeek moment” before us. Maybe not as notable as last year’s, but equally relevant. In Xataka | DeepSeek promised them happiness as the great Chinese AI. I didn’t count on a small detail: Kimi

Baidu has just launched a new AI model that competes with the best. The surprise: it is not Open Source

by

Baidu has presented Ernie X1.1, a new generative model that represents a promising qualitative leap and that seems to compete from you to you with their rivals. That is not so surprising. What it is is that we are facing a model that goes against the Chinese trend: it is closed. Why is it important. Being “the Chinese Google” worked very well for Baidu for more than two decades, but in recent times this Chinese technological giant is trying to Do not lose commits in the AI ​​race. His last movement goes in that direction: Ernie X1.1 has just presented, the latest version of a reasoning model that earns whole. The model can be easily tested in the Official Chatbot Website (You have to display the option to use the Ernie X1.1 model if you are not selected). The new Ernie X1.1 competes from you to you with Depseek-R1, Gemini-2.5 Pro and even GPT-5 in various benchmarks Promising. The internal tests published by BAIDU reveal that their new reasoning model offers “significant advances” in terms of precision of the answers (34.8% better than Ernie X1), instructions monitoring (12.5% ​​better) and agricultural capacities (9.6% better). Not only that: its performance in Benchmarks exceeds Desepseek R1-0528, the reference AI model for months in China. It is also able to compete in performance with GPT-5 and Gemini 2.5 Pro, two of the most advanced models today. But closed. While the trend in China is to offer open models in which at least the weights with which these models have trained, Baidu has decided that at the moment Ernie X1.1 will be closed. The model has been offered to corporate clients already developers through the Baidu cloud platform, called Qianfan (via API), in addition to its aforementioned availability in Ernie.baidu.com. Meanwhile, other models open. What it has done is to open the Ernie-4.5-21B-A3B-Thinking model, a version that is somewhat less capable but that has reasoning capacity and that supports a context window of 128,000 tokens. This model uses a MOE (Mixture of Experts) architecture with 21,000 million total parameters although only 3,000 million of them are active to make it more efficient. The model is for example available Through Hugging Face. An AI that accompanies the elderly. As they point out In SCMPduring the presentation of this wang haifeng model, Cto de Baidu, advertisement The launch of an AI agent destined to accompany the elderly. The objective is to help the more than 310 million Chinese who are over 60 years offering information that for example can help them with their health status and also allows us to discern whether the information they find on the Internet can be harmful to them. Counting Erres. In Xataka we have been able to use Ernie X1.1 for a few hours and the first impressions are very positive. The reasoning model is especially careful when answering, and for example when asking “how many” R “does the phrase ‘the San Roque dog has no tail because Ramón Ramírez has stolen it’?” The model correctly replied that 9, when Ernie 4.5, also available on that website, replied quickly but badly: 8. We also ask you to create a table with the 10 countries that had won the most football, and once again The answer It was perfect, something we had not seen in almost any previous model. In other interactions – for example, when creating and analyzing Python code – the performance was also remarkable. In the absence of trying it in depth, these small tests point to a really remarkable behavior of the model. But rivals squeeze. We are therefore before a striking step by Baidu, both for that promising performance and for the fact that the company has given up not opening it for the moment. In front of Ernie X1.1, yes, the competition is fierce. Both QWen3-Max-Preview (Alibaba) and Kimi-K2-0905 (MoNshot) have achieved enter the list Of the 10 most powerful models on the market According to LMarena. There for those who lead are Gemini-2.5 Pro, Claude Opus 4.1, O3 and GPT-5-High, but the advance of these Chinese models is remarkable. It remains to be seen if Ernie X1.1 will also sneak that list. In Xataka | Deepseek has given the starting gun in the race for a cheaper AI. And China starts with advantage

This exclusive Netflix thriller competes in intensity and brutality with ‘John Wick’ and is already number 1 in 27 countries

by

The most hardcore fans of action cinema can become as demanding as those of the most cinephile genres: That is why the choreographies, the edition and the rhythm of any new film that premieres of their favorite style examine the millimeter. And that’s why films such as ‘Kill Bill’, ‘The Raid’ or ‘John Wick’ are contributed: it is not easy to satisfy them. I may Netflix has found the key To do so with ‘Oni-Goroshi: City of demons‘, a frantic and violent epic of extreme action that has perched on the number 1 of platform views in 27 countries. The reasons are clear: a clear and linear argument that gives rise to innumerable sequences of action shot and edited with a taste that we do not usually see very often in the exclusive of streaming. The film is based on the manga of the same name as Masamichi Kawabe, and introduces us to a retired hired killer who lives with his family. But When your last mission fulfills, the consequences are unpredictable: A group of criminals murders his family. Given dead, he will undertake a devastating revenge 12 years after the tragic facts. And although the result does not reach the category of total classics of the genre commented above, the result is brutal and frantic enough (especially in its initial stretch, when it adopts a more realistic and dirty perspective) to justify the tremendous success it is having. In times of scarcity, films like this ‘Oni-Goroshi’ show that the quality action cinema is far from dying. Header | Netflix In Xataka | The director of Netflix’s best action film returns with a new lead and blood epic at the height of his legend

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.