Chinese

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

35% of its chip manufacturing machines are already of Chinese origin

by

Foreign lithography and wafer processing equipment manufacturers are selling less and less in China. In 2024, the country led by Xi Jinping represented 41% of ASML revenuebut in 2025 this figure dropped to 33%. And presumably in 2026 will contract up to 20%. Something very similar has happened to the American wafer processing machine manufacturer Applied Materials: its sales in China have gone from 37% of its total sales in 2024 to 30% in 2025. In addition, sales in China of the American companies Lam Research and KLA, and the Japanese Tokyo Electron, also have decreased during 2025 compared to those they obtained in 2024. This obvious trend is the consequence of two factors. On the one hand, US sanctions prevent US and allied manufacturers of lithography and wafer processing equipment from delivering their most sophisticated machines to their Chinese clients. The Dutch company ASML is most likely the most affected in this scenario. On the other hand, in response to pressure from the US, the Chinese Government is supporting the adoption of machines of Chinese origin in its integrated circuit factories. In fact, in 2025 the national tools represented 35% of the equipment in use in semiconductor plants, and Xi Jinping’s Government aims to reach 50% in new factories during 2026. Its purpose is clear: China’s chip industry needs to achieve technological independence as soon as possible in its fight with the United States. China has made great progress, but lithography remains its weakest point The resources that the Chinese Government is allocating to its designers and manufacturers of wafer processing equipment are bearing fruit. And they already compete face to face with foreign companies in the field of deposition, thermal processing, etching and cleaning of wafers. However, there are still no extreme ultraviolet (EUV) photolithography machines of Chinese origin in Chinese IC factories. Presumably they will arrive before this decade endsbut this is for the moment China’s real Achilles heel. One of the Chinese companies worth keeping track of is Pulin Technology. This organization has opted, like Naura Technology, AMEC (Advanced Micro-Fabrication Equipment Inc. China) or Piotech Inc., to develop your own cutting-edge photolithography machines. And the achievements are coming little by little. In mid-2025 Pulin sent one of his clients your first cutting-edge equipment using nanoimprint lithography technology (known as NIL for its English name NanoImprint Lithography). In mid-2025, Pulin sent one of its clients its first cutting-edge equipment NIL technology is not new. The Japanese company Canon has its own commercial NIL solution for yearsand presumably its operating principles are essentially the same as those of the machine designed by Pulin. On paper, NIL photolithography equipment is an alternative to printing machines. extreme ultraviolet lithography (UVE) designed and manufactured by the Dutch company ASML, although no to the high aperture version of these teams. The latter are currently the most sophisticated and expensive that exist. Very broadly speaking, the production of silicon wafers in the latter requires very precisely transporting the geometric pattern described by the mask to the surface of the silicon wafer using ultraviolet light and extremely refined optical elements. NIL lithography, however, allows the pattern to be transferred to the wafer without the need for intervention in the process. an extremely complex optical system. This strategy is simpler and cheaper, but it also involves the execution of several sequential processes that make it slower than UVE and UVP lithography. Canon assures that its nanoimprint lithography equipment can be used to manufacture integrated circuits comparable to the 5nm chips that TSMC, Samsung or Intel produce with ASML’s UVE machines. And in the future, with the refinements that will arrive, they will be able to manufacture 2nm chips. In addition, a NIL equipment costs ten times less than an ASML EUV machine: 15 million dollars compared to the 150 million dollars that the Dutch company asks its clients for an EUV machine with numerical aperture 0.33. We still don’t know how much each Pulin NIL machine costs, but it is reasonable to predict that at most it will have a cost comparable to that of the Canon machine. Image | Naura Technology More information | Tom’s Hardware In Xataka | Japan wants to end the Netherlands’ leadership in lithography equipment. This is your plan to get it

Meta spent 2 billion on a Chinese AI startup. China is clear that it was a conspiracy

by

With China and the United States dancing the dance of artificial intelligenceboth countries and companies want to get the best cards for their decks. Meta is investing millions in the development of AI and, even so, it seems to be lagging behind. To turn the tables, he closed 2025 with a $2 billion purchase: that of a Chinese startup called Manus. The operation was so notorious that the Chinese government itself raised an eyebrow and undertook an investigation to see what was happening there. And they are already clear. It was a conspiracy. The Manus case. Although it has rained a lot and these last few months in China AI companies have come out from under the stones, during the first half of 2025 the proper name was that of deepseek. It was the great competition from the Western OpenAI or Google Gemini, but in March something that looked like an AI agent began to appear: Manus. That’s how they sold italthough it was really a deep investigation mode that helps you perform actions, but does not do them for you. It didn’t matter: the expectation was there and, although there were doubts about his behavior and limits, Manus began to move a lot of money (more than 100 million in estimated income) and attract attention from the big players. One of them was Meta, who took over the company. The purchase. A good question is how China let something like this slip away for a technological and strategic rival to buy. And it’s a good question, but the answer is that, at some point, Manus stopped being a Chinese startup. In the middle of last year, Manus moved to Singapore, allowing the company to bypass export and import controls imposed on China. To the not having your own LLMthey depended on others like Claude which they could more easily access from outside China. This already set off alarm bells in the Government, but with the purchase of Meta the bells echoed. China put to work to various organizations to see what was really happening, the largest of them being the Chinese National Security Commission, which is commanded by President Xi Jinping himself. The reports prepared by this body are directly supervised by the leaders of the Communist Party, so it is a voice that must be taken into account. Conspiracy. And the result of the investigation is clear. As they comment in Financial Timesthe conclusion is that Meta’s acquisition of Manus is a conspiratorial attempt to try to undermine China’s technological capabilities. These are big words that do not remain in a vacuum, since the founders of Manus – Xiao Hong and Ji Yichao – were summoned by the NDRC last March to address issues such as possible violations of foreign investment rules in China. He did not stay for a meeting and, as the FT points out, both have been prohibited from leaving the country during the review process. In fact, there are sources that suggest that Manus would be considering backing out of the agreement, but even so, it is not clear that the Chinese authorities will be satisfied. For his part, Meta points out that they did everything according to the law and it seems that he has already started to integrate Manus systems into their tools, so taking that step back would be very complex. And now… what. That the National Security Commission has classified the case as “conspiracy” is something serious, since it was the trigger for a broader review that involves more agencies in the country that are currently reviewing everything. And the underlying problem is the speed with which everything happened. Manus took off and, just four months later, they moved everything to Singapore to break away from China just before the purchase of an American company. The investigation is shaking the Chinese technology sector because it is not the first time something like this has happened. Although on a smaller scale, it is an operation called ‘Singapore washing’ in which startups founded by Chinese move to the city-state to bypass China’s control and have a more direct line with the United States. The problem is that, at a time when the commercial and strategic war has intensified, calling the Manus case a “conspiracy” sets a precedent. One in which it is stated that China does not want to let artificial intelligence talent and technology escape because this advance has become one of the country’s strategic legs for the next five years. We will see what happens when the case is resolved, but it is clear that Beijing’s objective, like Washington’s, is to prevent its assets from escaping, and Manus can be the example for national technology companies do not follow a similar model in the future. In Xataka | We don’t know if “crisis” means “opportunity” in China, but there is one business where it does: RAM memory

Now a Chinese giant wants to turn it into the storage capital

by

Navarra is a consolidated world leader in renewable energy, especially a leader in wind energy: back in 1994, it built its first park and since then it has not only dressed some of its most iconic mountains with wind turbines, but has also created a powerful business ecosystem around them that has earned it the nickname of being the Silicon Valley of wind turbines. The starting point. This industrial and technological tradition has laid the necessary foundations to be a pioneer in the next step of the energy transition. After almost three years of contacts and visits, Navarra has closed the agreement to install a Hithium Energy Storage battery gigafactory. As summarized the president of the Foral Community: “We know that China has decided to go out and invest in Europe and these are opportunities that Navarra cannot miss.” The project. The Government of Navarra together with Hithium Energy Storage will form a joint venture through the public company SODENA for the construction of a plant that will manufacture lithium iron phosphate (LiFePO4) cells and assemble complete stationary storage system batteries (BESS batteries). The agreement signed in China contemplates an investment of 450 million euros, according to Qui Tangdirector of Strategy at Hithium, and the creation of around 700 jobs, with the possibility of a second phase that would add about 300 more, as collects News from Navarra. Still with some to be closed, the start of production has been set for 2027 and the location with the most votes for its installation is the old BSH plant in Esquíroz/Ezkirotz, just five kilometers south of the Navarrese capital. Why is it important. The great pending issue for the European energy transition involves stationary batteries: without the capacity to store electricity generated by the sun and wind, the electrical system cannot absorb more renewables without compromising the stability of the network. Europe has a lot of installed generation and little of its own storage industry and that gap is precisely what Hithium comes to solve. On the other hand, manufacturing within the EU allows it to avoid tariffs and other bureaucracy that Brussels applies to Chinese energy products. For Navarra the impact has several levels. Beyond the obvious economic investment and employment, it represents a logical evolution: from becoming a benchmark in clean energy generation to becoming the manufacturing hub for the storage systems that will allow it to be managed. But Hithium will also be the second battery factory in Navarra: Hyundai Mobis already has its own in Noainalthough oriented towards electric vehicles. With this arrival, Navarra consolidates an industrial ecosystem around energy storage at the level of few European regions. Context. The movement comes at a time of reconfiguration of European industrial sovereignty against third parties through the Net-Zero Industry Act (NZIA). That Hithium chooses Navarra makes it possible for its production to carry the “Made in EU” label, thus complying with the sustainability and regulatory standards of the old continent. Navarra already had installed power and leading engineering companies, but it lacked the capacity to store energy. In the midst of the global energy crisis, demand for large-scale storage solutions has skyrocketed and Hithium was looking for a strategic port to supply the European market. The Foral Community has forged this agreement slowly, as the counselor explains of Industry, Ecological and Business Digital Transition of the Government of Navarra Mikel Irujo, after three years of negotiations and seven visits. In fact, Navarra has carried out a strategic project that was initially aimed at Euskadi, as ElDiario.es points out. Who is Hithium?. It may not be as well-known as CATL, BYD or Huawei, but Hithium Energy Storage Technology is a giant in its segment and is also a real teenager. Founded in 2019 in Xiamen (Fujian Province) by Wu Zuyu, engineer specialized in batteries and ex-CATLthe world’s largest manufacturer in the sector. Of course, unlike CATL, whose main business is batteries for electric vehicles, Hithium focuses exclusively in stationary storage for the electrical grid. Less media coverage than electric cars but essential for the energy transition. In record time she has gone from being unknown to becoming one of the leading manufacturers global stationary storage battery companies with a presence in more than 20 countries and more than 1,200 engineering professionals in a workforce of 8,000 people. The one in Navarra will be Hithium’s second factory outside of China (the other is in Dallas, Texas) and will be the company’s reference base for all of Europe. Yes, but. Although the agreement has been signed, Chivite itself has warned that there are still administrative steps pending and the location is still unclear. On the other hand, the joint venture with the participation of SODENA implies that the risk does not fall solely on Hithium: if the project is delayed or does not reach the objectives, the Navarrese citizens assume part of the consequences. The commitment to local employment is explicit and the Provincial government has promised to monitor itbut there are already precedents that invite caution. Leaving aside this specific case, there is a technological question that remains unanswered: Europe has foreign dependence on rare earths or semiconductors and this risk also affects batteries. It is true that a Chinese gigafactory on European soil generates local employment and produces within the EU, but the technology, intellectual property and strategic decisions continue to be made in China. In Xataka | The solar miracle that went wrong: Spain produces more electricity than it can manage In Xataka | We have a problem with heat in buildings. A Navarrese investigation knows how to cool them without air conditioning Cover | Hornsdale Power Reserve and Pamplona City Council

Chinese companies are designing their AI for an audience the West is ignoring: retirees

by

We can adapt the title of that great Cohen film to AI: there is no AI for old people. The majority of AI chatbot users are young and the older ones usually have technical knowledge. The AI ​​boom, like other technological booms, is leaving out the older onesexcept in China. Hello, grandmother. They tell it in Nikkei Asia. Large Chinese technology companies such as ByteDance and Tencent are designing chatbots and apps with AI with older users in mind. Doubao, the most popular chatbot in China, has launched advertising campaigns targeting retirees, highlighting its accessible features; It allows you to converse by voice, understands dialects and even addresses users as grandfather or grandmother. According to data from China Internet Network Information Centerthe number of AI users between 50 and 59 years old represent 10% of the total and those over 60 years old only 5%. They are still a minority, but there is a curious fact and that is that, although the adoption rate in this group is much lower, the users who start using it are more loyal and use it frequently. Everyday help. In Nikkei they tell the story of Chen Bing, a 63-year-old woman who has made AI her personal assistant. He used it to organize an event with alumni of his school, from sharing expenses to generating a video that he used in the background of a poetry workshop. It also helps you identify flowers and read fine print. According to Chen, AI gives him independence and prevents him from having to constantly ask his children for help. And health. There are other AI proposals aimed at the elderly, such as Ant Afu, a health chatbot with which users can get advice and access health services. However, it has generated criticism, first of all due to possible conflicts of interest. In the past, there was a scandal because Baidu recommended hospitals and treatments based on paid advertisements and there are doubts that this system has similar influences. On the other hand, there is the question that AI continues to fail a lot in diagnosis. The silver economy. It is what the market for products and services aimed at older people is called in China. China already has 323 million retirees and the government is promoting these types of initiatives since it sees great potential for consumption by the elderly, something they need to encourage in the midst of an economic recession. It is estimated that by 2035, the silver economy will account for 10% of the country’s entire gross domestic product Aging population. It is one of the problems facing China today. The government is trying literally everything for stimulate birth (without much successby the way) and have also raised raise the retirement age. However, the aging of the population is not something exclusive to China, it is also a problem that Europe and more countries in the northern hemisphere We have been dragging on for a long time. In the European Union there are some initiatives such as digital literacy courses for seniorsbut at the private company level, the proposals are very niche. In Xataka | China knows that its population is going to collapse but it already has a long-term plan to solve it. Of course, thanks to AI

EEEU vetoed the largest Chinese drone manufacturer. He did not expect that he would be left without the largest Chinese drone manufacturer

by

In December 2025, the US government banned DJI, making it the Huawei of drones. It was an absolutely crazy idea.with American drone pilots themselves warning about the Trump administration’s terrible decision. To no one’s surprise, the play did not go well. what happened. Late last year, the United States Federal Communications Commission (FCC) decided ban all drones and critical components of these small aircraft that were manufactured in foreign countries. The measure affected the import of new drones, remaining existing ones operational. But the government did not take into account a small detail: DJI is the main reference in drones worldwide and, literally, there are no alternatives. What is happening. Already in 2025, Greg Reverdiau, co-founder of the Pilot Institute in Arizona, conducted a survey in which 8,000 pilots participated. 85% made it clear that they could stay in business for about two years. From then on, without access to DJI drones, the outlook was unsustainable. Photographers Videographers Farmers Surveyors Emergency services Security forces Major figures in the industry make it clear that no one is going to replace the gap that DJI has left, whether in capacity, affordability, reliability or ease of use. The alternative. GoPro launched a drone, Karma, in 2018. It failed and was never heard from again. Companies like Parrot also launched consumer drones almost ten years ago, but today there is no trace of them. American companies like Skydio have pivoted completely towards defense, with drones worth thousands of dollars and million-dollar contracts with the US military. When asked if they intended to manufacture drones for consumption, the answer was a clear no. Goodbye to 90%. DJI dominated the US drone market with a 90% share, and there are no real alternatives to replace drones that are reaching the end of their useful life. With no possible DJI replacement in sight, the question is no longer who will take over, it’s how long the current fleet will last before volunteer firefighters, farmers and rescue teams run out of work tools. In Xataka | Best drones. Which one to buy and recommended models from 50 to 3,500 euros

After visiting a Chinese factory, the CEO of Honda loudly admitted the noise of the industry

by

We are witnessing a great change in the automobile industry, led above all by the great presence of China in more and more global markets and a transition to electric which seems to still be difficult for him. The traditional automobile industry is going through a delicate point, and the president of Honda saw it clearly when visiting a supplier factory in Shanghai. The surprise. At the end of February, Toshihiro Mibe, president of Honda, visited the facilities of a large Chinese manufacturer of components in Shanghai. What he found was a completely automated plant, without workers on the production line, and capable of supplying parts to both Tesla and local builders, minimizing labor costs and operating constantly. “We have no chance against this,” counted Mibe when leaving, according to statements reported by the Nikkei Asia media. It is certainly not the type of statement that one would expect from someone who runs one of the most historic brands in world motorsport. Why does it matter? Honda is not an isolated case. It is the latest symptom of an industry that has been looking at China with concern for years. Chinese manufacturers have managed to compress the development time of a new model to between 18 and 24 monthsabout half of what the Japanese or Europeans need. And it’s not just speed: it’s cost, automation and software. It is a change that is costing the traditional automobile industry, and that is not easy to replicate either. Numbers. In 2020, Honda sold 1.62 million vehicles in China. In 2025, that figure fell to 640,000 units, a decrease of 24% in the last year alone and the fifth consecutive year of decline, according to data published by the media. Its factories in the country operate at 50-60% capacity, well below the 70-80% necessary to be profitable. By 2026, the planned production is less than 600,000 units. “It is an extremely disappointing plan,” acknowledged an executive from a Chinese supplier company to Nikkei Asia. “But it doesn’t surprise me either,” he continued. Honda is not alone in this. Jim Farley, CEO of Ford, warned in an interview with CBS Sunday Morning last October that China has enough production capacity to “supply the entire North American market and put us all out of business.” “Unless things change, we will not survive,” counted for his part, also the then president of Toyota, Koji Sato. And coming from Toyota, which is basically the largest automaker in the world, that says a lot. Vgo back to the past to go towards the future. Honda’s reaction goes through resurrect your R&D division as an autonomous entity, something that has already existed since 1960 and that in 2020 was dismantled in favor of centralized management. It was that independent structure that, in 1972, developed the low-emission CVCC engine (the first to meet US regulations) and turned the original Civic into a global success. Now, thousands of engineers return to a subsidiary with greater operational freedom. “Five or six years ago it was good for the headquarters to take the reins,” recognized a Honda executive to Nikkei Asia. “But now the world has changed drastically,” he continued. Doubts. The movement does not convince everyone. Takaki Nakanishi, chief analyst at the Nakanishi Research Institute, said to the media that “it is doubtful what will change just by restoring the organization.” Honda’s own management team admits that recovering the structure does not guarantee winning China. “But that doesn’t mean we’re going to raise the white flag,” added a company executive, according to Nikkei Asia. In parallel, Honda cancels two of its electric planned for the US, the 0 SUV and the 0 Sedan, and assumes losses of up to 15.8 billion dollars. Also have been left in the air the two vehicles under the Afeela brand, the joint project with Sony. The alternative bet: India. While Toyota and Nissan choose to ally with Chinese partners to learn from their speed and launch affordable electric cars, Honda prefers another path. The brand is betting on India as a manufacturing base for its next generation of electric cars. The Model 0 Alpha, its global strategic EV planned for 2027, will be produced there. In mid-March, the Indian subsidiary shared images of the Alpha in rolling tests, describing the moment as “a new milestone in Honda’s electrification journey.” Imbalance. The automobile sector is going through one of its most profound transformations. China has stopped being just a market to become the main global competitor, with brands like BYD already reaching 1.8% share in Europe in the first two months of 2026, according to data from the European Automobile Manufacturers Association (ACEA). Honda, with just 0.5% in the same period, illustrates this imbalance well. Cover image | Sling In Xataka | Sensors, luminous tires and fish scales: the crazy (and stinky) story of the first “autonomous” car

Chinese AI models boasted of being good, pretty and cheap. There are only two of those three things

by

It is not as well known as its rivals, but Zhipu AI (z.ai) has become one of the most promising Chinese AI startups. It is responsible for the family of open GLM models that have always offered a solvent and, above all, very cheap alternative. That, unfortunately, is no longer so true, but we are witnessing a change in strategy both between it and its competitors in the Asian giant. Chinese AI models are no longer such a bargain. GLM-5.1 is better… Z.ai announced yesterday the launch of its shiny new AI model, GLM-5.1. I did it with my chest out because we are facing a promising evolution of this LLM (744B parameters, 40B assets with Mixture of Experts architecture) that certainly surpasses its predecessors but that in some metrics even seems to be above GPT-5.4, Claude Opus 4.6 or Gemini 3.1. Agentic tasks and those that require autonomy for long periods work better than ever, but if you want to benefit from these improvements, you have to check out: the price of the model is now at least 8% more expensive than previous versions. …but also more expensive. According to prices managed by OpenRouter, the well-known platform that serves as a “distributor” of multiple free and commercial models, the prices of the new Z.ai model have risen significantly. Thus, GLM-5.1 costs between 8 and 17% more than GLM-5 Turbo, also recently launched. It is the second time that the Chinese company has raised prices for its users in 2026, and that is a worrying sign. The excuse, of course, is the same as always. We are in high demand. When Z.ai launched GLM-5 at the beginning of February, it took the opportunity to raise the prices of its plans for programmers between 30 and 60%while the API rose between 67% and 100% (doubling). Its shares on the stock market perked up significantly after the launch and the price increase – logical, investors saw that income was probably going to increase thanks to these increases – but the company indicated that demand was very high and that its models had to reflect that circumstance. From the three B’s to just two. The Chinese open models had been demonstrating remarkable quality and a fantastic price/performance ratio for months. They were good, pretty and cheap, but Zhipu AI has just been the latest to end up raising prices. Most of its competitors have been doing it too: Moonshot AI (Kimi), MiniMax and StepFun did it already in 2025, but Alibaba, ByteDance, Tencent and Baidu have also adopted increasingly ambitious pricing strategies. as indicated on TrendForce. OpenClaw as a trigger. Much of the blame for this great demand lies with AI agents like OpenClaw, which has become viral but has a problem: it consumes tokens at an extraordinary rate. A conversation with ChatGPT, Claude or Gemini has a cost, but the use of tokens in “chat mode” is much lower than that carried out by AI agents, who do not stop “thinking” and analyzing different possibilities and chaining processes to resolve our requests. The Chinese models have become a good alternative if one wants to save because using Claude Opus 4.6 was very expensive —and now, prohibited—, but these models are slowly becoming high-end AI models. At least, for price. I already know how this story ends. What we are experiencing with AI models we already saw with smartphones. Chinese manufacturers broke the market with bargain phones that offered high-end features for mid-range or low-range prices, but then they evolved and over the years most manufacturers have ended up focusing on the super-high ranges and at most have launched “cheap” sub-brands. Xiaomi has done it with Redmi and POCO, for example, and now we are seeing something similar with Chinese AI startups, which gained popularity with good, pretty and cheap models, but are now beginning to transition to that new batch of capable but no longer so affordable models. First they catch you, then they squeeze you. What we are seeing with the Chinese AI models we were also seeing with the models of companies like OpenAI or Anthropic. Both they and their competitors release increasingly better but also increasingly more expensive models, and that means that those tokens that these companies sell us are becoming more and more precious: the quotas for the ChatGPT Plus or Claude Pro plans, for example, seem to be running out. faster than beforeand the users they take time complaining about it. On Reddit They have a “megathread” dedicated precisely to that, but here we have bad news: this doesn’t look like it will go down, but rather more. In Xataka | Anthropic has shut down OpenClaw for a reason: it’s building the “walled garden” that Nintendo perfected

DeepSeek promised them happiness as the great Chinese AI. I didn’t count on a small detail: Kimi

by

Just a year ago, DeepSeek was one of the biggest scares that Silicon Valley had received dwarves. A Chinese model trained with a fraction of OpenAI’s budget equal to GPT-4 in benchmarks. Upon its arrival the message seemed clear: Western dominance of AI had its days numbered. Today, the story stands, but not thanks to DeepSeek. The DeepSeek case. DeepSeek carries months late for its V4 and, to date, has already lost three of the authors of R1, the model that catapulted them to success. The monthly downloads fell 72% in the second quarter of the year, seeing how Doubao (ByteDanec) snatched the lead. With missed dates, usage errors due to cyber attacksand the difficulty of split from NVIDIA To bet almost entirely on Huawei’s Ascend chips, Chinese alternatives like Kimi have been gaining ground. Meanwhile, on the other side of China. Moonshot AI was not born surrounded by noise like DeepSeek. It was founded in March 2023 by three former colleagues from Tsinghua University: Yang Zhilin—PhD from Carnegie Mellon, former Google Brain and Meta AI—, along with Zhou Xinyu and Wu Yuxin. There were no visible or media faces behind it, only product. That product is Kimi, and in early January 2026 the company launched it in its K2.5 version. In code and video benchmarks managed to surpass GPT-5 and Gemini Pro 3with the key to Chinese AI: its API costs between 4 and 17 times less than OpenAI’s. Those responsible for Moonshot explained how Kimi was almost at Claude’s level in software development testing, encouraging the race for open models. The money arrived. The commercial results are what really attract attention. In less than 20 days Following the launch of K2.5, Kimi’s cumulative revenue exceeded everything billed during 2025. API’s international revenue increased fourfold since November of the previous year. The consequence in valuation has been dizzying: 4.3 billion dollars in December 2025, 10 billion in February 2026, 18 billion in March. Three months, valuation multiplied by four. Kimi has thus become the fastest decacorn in Chinese business history. The Chinese maelstrom. DeepSeek was born a year ago as the great revolution that questioned the closed model of Silicon Valley. It only took a few months for Moonshot to steal the limelight and manage to be on par with – or even above – giants like Google and OpenAI in the most used models in the world. In favor of DeepSeek, it should be noted that its objective is different: it does not follow the typical startup pattern with pressure for immediate monetization and it is a gigantic AI laboratory that can afford not to win in the short term. In Xataka | DeepSeek API: what it is, what it is for, prices and how you can get one to use in your projects

I thought Chinese cars were going to be the new Android. They are actually the new iPhone

by

For a few years, the nightmare of European manufacturers has had a specific name: the Android scenario. That Google, or Apple, or Amazon, would turn the car into interchangeable hardware. that the value will migrate to third-party software and they will be reduced to outdated manufacturerslike PC manufacturers in the nineties. That fear has kept them on guard, looking towards Silicon Valley, investing a lot of money in their own connectivity and infotainment systems, trying not to be left out. They’ve been guarding the wrong door. The movement that is happening is not the equivalent of Android. It’s exactly the opposite, at least where it hurts the most: BYD makes its own batteries, your own operating systemand operates its own charging network. Xiaomi does practically the same with HyperOS. The logic is not to create a platform where others monetize but to control every centimeter of the experiencewithout intermediaries. That has a name that we all recognize, and it is not Google. It’s Apple’s. The paradox is that the Android scenario that Europeans feared so much is happening, but it is not being carried out by the big American technology companies, they are building it themselves: European generalists have become what they feared most, without anyone from outside having to impose it on them. What makes the Chinese movement so different is what is noticeable inside the car. Denza, YangWang, Luxeed, Exeed either Xpeng They are brands that three years ago almost no one in Europe even knew about, but that Today they are manufacturing cars with interiors with an attention to detail that is very reminiscent of what happened with the iPhone in 2007.. It wasn’t that the iPhone did more things than the competition (at the time, in fact, it did considerably less than a Nokia). It was that every interaction was thought out, every transition animated, every small gesture had coherence. Rivals had cool features, but Apple had experience that no one matched. Today, sitting in a mid-range or high-end Chinese car and sitting in a German car of the same price is not so much about comparing specifications as it is about comparing philosophies. And the Germans, who are seeing it, are reacting: the new iX3he CLAeither the newly announced i3 They are serious efforts to recover that coherence of experience. But reacting is not the same as taking the initiative. The problem facing the European industry is not that it does not know how to make cars, there is more to it. The thing is that for many years the margin has been captured by those who mastered mechanical engineering, and they learned to optimize exactly that. What they did not learn is that in the 21st century the margin is captured by whoever controls the entire experience: the software, the data, the services, the ecosystem. When they wanted to learn it, they looked at Silicon Valley because there was the model they knew. Until four days ago, no one looked towards Shenzhen, where someone had spent years building something more like Apple than Google: vertical, closed, cohesive, with a speed of iteration that Westerners simply do not have and they already admit it. Nokia also had very good engineers. In Xataka | At 110 km/h and driving every other day: Europe already has its recommendations for the latest oil crisis Featured image | BYD

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.