Chinese

Making a robotic hand is “100 times more difficult” than building the entire robot. These Chinese companies are determined to solve it

by

China has achieved let your humanoid robots run, they fight and even that execute choreographies with astonishing precision. That was the easy part. The key for a humanoid to stop being a fairground attraction and become a useful product is not in its legs, balance or its “brain”, but in something much smaller and much more complicated than it seems: the hands. The challenge of the hands. The human hand has 27 bones, 34 muscles and countless nerve endings. Replicating this in metal and circuits is the biggest bottleneck facing modern robotics. sums it up well Guardian Zhou Yong, founder of LinkerBot, one of the most advanced Chinese startups in this field: manufacturing a robotic hand is “a hundred times more difficult” than manufacturing an entire humanoid. “The dexterity required by a hand is ten times greater than that of any other part of the body, but its volume is only one-tenth the volume,” Zhou says. That is, not only is it the most technically complex element, it is also the smallest. The Chinese advantage. Pan Yunzhe, founder of Wuji Technology, studied in the United States and considered setting up his robotic hands company there, but soon saw that it was unviable. He tells The Guardian that “It was practically impossible to manufacture hardware in the United States due to the enormous limitations of the supply chain.” And China has an unmatched supply chain: it is agile, it is sophisticated and, above all, it is cheaper. This gives them a key advantage in hardware and allows companies like Linkerbot to already manufacture 5,000 robotic hands per month, a figure impossible to match anywhere else in the world. The problem is the software. Making the hand is only half the job, you also have to make it move like a human hand and there is still a long way to go here. Nathan Lepora, professor of robotics and AI at the University of Bristol, sums it up: “the challenge of making these hands is already being solved,” but controlling them “is a completely different game… no one knows how to do it yet.” The scarcity of data is one of the big problems and it is that, while LLMs have been trained with the enormous amounts of data on the Internet, there is hardly any data on how a human hand moves and, above all, what it feels like when touching something. To try to overcome this obstacle, Wuji Technology is testing a glove packed with sensors that captures all the movements of the human hand in everyday tasks. Its founder admits that being able to capture “how a person moves and what they touch or feel” is an “extremely complex and unsolved” task. The robotics market. The Chinese robotic hands sector had a turnover of $7.4 billion last year, almost four times more than in 2024. LinkerBot, one of the leading startups, aspires to a valuation of $6 billion. The case of this startup is just one example of the robotics boom in China, where more than a million companies are already registered, 40% more than last year. This translates into brutal dominance: China already manufactures 90% of the world’s humanoid robots and it is the country with the most industrial robotsby much difference with the rest. Yes, but. Robotics is advancing by leaps and bounds, but there are voices calling for calm, and not just any voice, but that of the International Federation of Robotics. In your report published in September last year were clear: “True multipurpose humanoids are still far away.” We are going to continue seeing more and more amazing demonstrations, but from there until they are sold en masse and we can all have one at home that is truly capable, there is a way to go. Image | Xataka with Magnific In Xataka | We still don’t know if humanoid robots will be the next great technological revolution. Yes we know that China will lead it

Apple is knocking on Chinese doors looking for RAM. According to the Bank of America, it is pure theater

by

Apple is in the same adventure as the rest of the companies that sell consumer technology: the adventure of getting RAM memory. Although during the beginning of the RAM and storage crisis it seemed that Apple was the only one that could endure the blows, reality ended up hittingdemonstrating that if they had not raised prices it was because they had plenty of stock in advance. When that stock disappeared, we saw reality: model withdrawals, price increases and, more recently, new price increases with a whopping 20% ​​more on iPad and Mac. The problem for those from Cupertino is that the key moment of the year is approaching: the launch of the iPhone 18 Pro and the rumored foldable iPhone, and there the price is going to be crucial. That’s why they find themselves looking for RAM and storage even under the rocks, with China being the market in which they have put the magnifying glass to see if memory from manufacturers like CXMT can fit your products. But of course, there is a problem: the United States and, above all, the Bank of America, which says that All this about Apple looking for RAM in China is a little theater. Teatrillo to get better prices A few days ago we said that Apple was knocking on doors. If Samsung, SK Hynix and Micron don’t open because they are so busy neglecting the consumer segment in favor of AI, then we have to ring new bells. With the cushion on the latest and new key products on the starting grid, Apple’s escape route seems to have a name: CXMT. ChangXin Memory Technologies is a Chinese DRAM manufacturer founded in 2016 that, since then, has been polishing its product lines, making significant advances to catch up with its large South Korean competitors in different DRAM models and technologies. That said, it is estimated that Apple is striving to bring positions closer together with CXMT to ensure supply, but this has caused Bank of America to raise an eyebrow. And even worse: CXMT is on a certain list of the US Department of Commerce. In a recent report, Bank of America analyzes the market situation, exposing that there are misconceptions about the memory chip sector and stopping, curiously, at Apple’s intention to approach the Chinese manufacturer. The bank sees positive the continuous and massive spending of Big Tech to make data centers, but when it comes to buying memory from China, they are not so positive. According to the bank, Apple’s recent efforts are just a movement, a little theater, to gain bargaining power against the usual suppliers. This makes sense. Apple is no longer Nvidia, but it is still a very big fish in the consumer sector and buys a huge amount of components from different manufacturers. If Apple reached an agreement with CXMT, it would have an impact on the accounts of suppliers such as Samsung or SK Hynix. Because in the end, that’s what this is all about: that the big fish has much more decision-making and negotiating power than the small one. When Valve wants to do its Steam Machine, You should accept what they offer you. at the price they offer it and without question, but Apple is, in the field of hardware, much more important than Valve and would have some negotiating power. For Bank of America, it’s all about that strategy because they believe that the memory that the Chinese can create is, currently, valid only for more modest deviceslike a hypothetical iPhone 18e, not for the top of the range like the iPhone 18 Pro, the iPads with M processors or, above all, the Macs. And more important than the institution’s analysis is the fact that CXMT appears on the Pentagon’s blacklist of Chinese Military Companies. According to the Financial Times, Apple has been pressing to the Government to being able to buy memory from CXMT without consequencesbut if tomorrow, for whatever reason, the Chinese company enters the other Pentagon blacklist (in which American companies are prohibited from trading), Apple would be in serious trouble. In the end, no matter what happens, what is certain is that we, the users, will be the ones we end up paying the consequences. And products like the aforementioned Steam Machine, but also what is projected PS6low-end mobile phones, Raspberry Pi, nintendo switch 2 either the new xbox They are examples of what we can expect: prices -very- rising. Image | Laurenz Heymann (edited) In Xataka | In 2007, Steve Jobs went on stage with an iPhone that barely worked: he was saved by a script that did not allow even a detour

hundreds of Chinese boats preying on fishing grounds

by

There are maps that speak for themselves, like the ones has disclosed during the last few days the Ecologist Movement of Peru (MEP), an organization dedicated to the defense of the environment. Last week its managers hung up two satellite maps in which dozens and dozens of colored arrows can be seen crowded together in the Pacific, just off the coast of Mollendo. Each one of them, MEP complaintreveals the position of a Chinese ship that goes to the edges of Peru’s national waters in search of its squid schools. His presence already has put on guard to local fishermen. What has happened? That MEP has stirred up an old debate in Peru (and other nations of South America): the impact that the Chinese flag fleet has on the maritime resources of the region. On June 22, the organization published a satellite map online showing the concentration of dozens and dozens of ships just 220 nautical miles off the coast of Mollendo, south of Peru. According to the organization environmentalist, there are around “300 Chinese vessels” dedicated to “exploiting fishing outside the limits” of their national waters. Just one day later, on June 23, MEP returned to the fray with another satellite map that shows a long trail formed over the Pacific by ship marking points. “Satellite images from June 1 to 19 show that the Foreign Squid Fleet has completed its migration (north-south) along the edge of the Peruvian EEZ,” warned the entity. “Around 400 Chinese fishing vessels are concentrated 220 miles off the coast of Mollendo.” @ecocentristas This is what the fleet of around 300 Chinese fishing vessels that are concentrated 220 miles off the coast of Mollendo looks like 🇵🇪 🛰️These are China’s boats, exploiting fishing off the limits of the seas of Peru. 🇵🇪🦑🦑🦑 Ecological Movement of Peru ♬ original sound – Ecologist Movement of Peru – Ecologist Movement of Peru What is the problem? Basically the number and location of that large fleet of fishing vessels. If MEP is correct, these are hundreds of Asian ships mobilized right on the border of Exclusive Economic Zone (EEZ), a region in which the coastal States (in this case Peru) have sovereign jurisdiction, which affects, among other things, their natural resources. The EEZ usually extends 200 miles, which would put the Chinese fishing fleet almost on the edge. The problem is that, beyond the demarcations drawn in offices or the distances included in international treaties, the Peruvian EEZ is situated in a much larger context: the ecosystem of the humboldt currentmaking that region of the Pacific especially valuable for fishermen. The UN itself recognizes that it is one of the “most productive areas in the world”, although it has also been warning for years about the serious threat which involves both climate change and the overexploitation of its fishing resources. What impact does it have? The million dollar question. In 2025 the newspaper The Republic public a report in which he echoed several complaints from Peru’s artisanal fishermen: incursions by Chinese vessels into the EEZ, indiscriminate exploitation of resources and fear that squid schools would be depleted. “The Peruvian boats go out, but they don’t bring the amount they used to. The Chinese boats prey on the sea, our boats are small, everything is done by hand. On the other hand, they have machines that take the fish faster,” explained to the newspaper Alberto Sánchez, fisherman from Paita, from Lima. Sailors dedicated to the artisanal capture of Pucusana even have denounced the sighting of large vessels in the 200 miles of the Peruvian EEZ despite the fact that the fleets must transmit their position via satellite. @ecocentristas 🛰️ Satellite images from June 1 to 19, 2026 show that the Foreign Squid Fleet has completed its migration (north-south) along the edge of the Peruvian EEZ. Around 400 Chinese fishing vessels are concentrated 220 miles off the coast of Mollendo. 🇵🇪 @ecocentristas ♬ original sound – Ecologist Movement of Peru – Ecologist Movement of Peru Is it something new? No. In 2024 MEP already launched a similar complaint. He even shared a map showing the accumulation of Asian ships right on the border of the Peruvian EEZ. “Where is the foreign squid fleet located? How many ships are there? What type of vessels are they and how many are in Peruvian ports?” I questioned the organization. A few days ago, after its last complaint, the Peruvian Navy (MGP) carried out an exploration flight which confirmed that, at least today, the foreign fishing fleet operates outside the Peruvian maritime domain, 230 miles away. Does it only affect Peru? No. The debate regarding the presence (and impact) of foreign fishing vessels on the South American coast is not new and goes far beyond Peru, also extending to nations such as Chili. Infoae cites studies that estimate that in 2024, 1,359 vessels will operate in the 500 nautical miles located off the coast of Peru. Of them, 525almost 40%, were ships of Chinese origin, a figure that far exceeds those of other nationalities. MEP’s warning also comes just a few days after Sustainable Fisheries Partnesihp launched a statement resounding in which they warn of the importance of not overexploiting the region’s resources. Hence, among other things, it requires that any legislative change be supported by a “scientific basis”. In the specific case of Peru, the agency warns of the registration of around 2,000 new vessels “built outside the legal framework” at a time when “the fishery has captured 83.27% of the quota” planned for this year. Does context matter? Yes. And not only because of the warnings from environmentalists or the misgivings of the sector. Two years ago the organization The Outlaw Ocean published a report in which he warned that China’s fishing footprint goes far beyond its fishing grounds or Asian flag vessels. The country also operates in other waters of South America, Africa and the Pacific thanks to ‘flagging’, which basically consists of arranging for a ship to fly the … Read more

The Ferrari Luce has sold all its units in China. And that is the best sign for Chinese manufacturers

by

Ferrari Luce in China. With the Luce, of course. The Maranello electric car has posted the “all sold” message on the door of its dealerships. 88 units at almost $600,000. A very high figure for a Chinese market that has also lowered the prices of luxury vehicles. Despite everything, Ferrari is Ferrari. whatever they want. They can sell whatever they want. It is the conclusion defended by multiple analysts and pickup in Xataka during the presentation of Ferrari Lucethe company’s first electric car that created enormous controversy for its design that even surpassed that of having launched an electric car without a trace of gasoline smell. Just a few days later we learned that Ferrari had sold each and every one of the units that I had planned in a first print run. Many or few? It really is not entirely clear because Ferrari’s production capacity is limited and only with the passage of years will we be able to know if the car has been a success or not. What is certain is that Ferrari, by the simple fact of being Ferrari, shows that it can sell whatever it wants. 88 units for China. With a waiting list that already extends until 2027, the Ferrari Luce has had the reception that could be expected. Even though the company has assured that does not force its more traditional clients to buy the car as a preliminary step to get the most special units. Of that waiting list that has already been completed, 88 of those units will go directly to the Chinese market. These Ferrari Luce, ensure in Car News Chinahave been sold with a starting price of $586,600 (almost four million Chinese yuan at the exchange rate). The figure is, they explain, 7% lower than the price sold in Europe. Very important. That sales are advancing at a good pace is huge news for the brand since China has been very quickly turning to its local manufacturers and turning its back on foreign ones. In fact, Ferrari only sold 584 units in 2025 in the Asian country. The fall was serious because in 2024 it signed 814 units and in 2023 it placed 1,221 supercars. That is to say, placing 88 units at once in China has already allowed it to sell 15% of the same amount that it signed last year. Without a doubt, a cushion that allows you to cushion the fall in a market that is clearly moving towards electrical products and, above all, to local ones. Big money. When analyzing car sales in China, in CarNewsChina They talk about “extra luxury” to refer to cars that cost more than a million yuan (just over 120,000 euros at direct exchange rate). It is a market where, curiously, only European cars dominate. The twentieth place last year was the Mercedes-AMG GLS with 83 units. In one fell swoop, Ferrari would have already surpassed it. But this list is not dominated by European cars because they do things better, it is because Chinese manufacturers have entered into a price war that has managed to place the price ceiling of their luxury cars below that million yuan. That is why it is surprising that Ferrari is capable of placing 88 units of a car that multiplies that cost by four. It must be taken into account that only BYD with the Yangwang U8 and its long variant U8L managed to sneak into this list. The rest of the supercars do not appear because, among other things, they cost less money than that million yuan border. Much more for less money. To show what we are talking about, the Huawei Maextro S800a luxury sedan that is half-manufactured by Huawei and the automobile conglomerate JAC, started last year at just over 700,000 Chinese yuan (about $100,000 at the exchange rate) and its top range barely exceeded that million yuan. Only in the last month of December 2025, This car outsold the BMW 7 Series, the Porsche Panamera and the Mercedes-Maybach S-Class. which were consolidated as the next three best-selling cars. While the Chinese car sold 4,376 units, the European options totaled 4,140 units. Chinese manufacturers are offering cars that are more technologically advanced inside than any other European company. The Huawei Maextro S800 has triple screen inside between 15 and 16 inchesendings in wood inside with space to store champagne bottles, individual seats in bucket format in the rear seats, the most advanced ADAS systems and, of course, full integration with the Huawei ecosystem. The path to choose. They assure Ferrari that the Luce is a car that is not aimed at its most classic clients. “It is designed for a different customer profile, not for historical customers to buy, who, obviously, can do so if they wish,” assured Enrico Galliera who has just left his position as the company’s head of marketing. As we said in Xataka It is clear that Ferrari is looking to carve out a niche for itself among a new type of customer. Although it aims at a global market, this was essential in China where the preferred car among the rich also seems to be electric and Chinese but, above all, it aims to be something different from what we traditionally know as a car. That Ferrari has sold all the units there allows it to position itself as a modern company, capable of doing different things. For example, this opinion from a Chinese buyer of a Porsche Taycan in statements to Bloomberg: “It was just an electrified Porsche. That’s all.” In the article, he described the purchase as “terrible” because, simply, it did not provide any differential value. For now, the numbers seem to say that Ferrari is moving on the right path in China. Photo | Ferrari and Sou Jest In Xataka | “It shouldn’t be in a car”: Legendary Apple designer thinks we have a problem with touch screens

Experts already claim that the Chinese GLM-5.2 model is as “dangerous” as Anthropic’s

by

The technological gap between the US and China continues to narrow. At least, if we pay attention to what they say the latest analyzes on the GLM-5.2 model. Two independent cybersecurity companies have made their own assessment and their data reveals that in terms of cybersecurity, GLM-5.2 is as good as Claude Opus 4.8. That has notable implications, especially considering how the US government is now restricting access to Anthropic and OpenAI’s frontier models. AI in the face of the threat of cybersecurity. Since Claude Mythos Preview appeared, the discourse on AI has changed significantly. Suddenly the world realized that these models could become weapons with which to find vulnerabilities in all types of systems to exploit them. Anthropic has already warned that Mythos was too dangerous to be publicly available, and it did not matter that it released hidden versions like Fable 5 shortly after: the US Government has temporarily vetoed them and the same has happened with GPT-5.6. The situation is unusual. Beware of Tulongfeng (or not). Last Wednesday, a Chinese cybersecurity company called 360 Security Technology (Qihoo 360) launched a new vulnerability detection tool called Tulongfeng. According to its creatorsTulongfeng is comparable Mythos in this task. The company this on the US “Entity List” since May 2020 and its CEO, Zhou Hongyi, stated that Mythos is equivalent to a “cybernuclear weapon.” The Sputnik moment with GLM-5.2. But the real recent protagonist of the Chinese AI industry is the GLM-5.2 model from the startup Zhipu.ai (Z.ai), which is becoming very popular by demonstrating performance comparable to the best models from US companies. Its fundamental advantage is that it is an open weights model: any person or company can download it, modify it and run it on their own hardware (although it requires a huge amount of video/unified memory to be able to use it, it is a model with 744B of parameters). But he is not only good at programming or at agentic tasks. Better than Claude in cybersecurity? The cybersecurity firm Semgrep stated in a recent analysis that GLM-5.2 was superior to Claude Opus 4.8 regarding cybersecurity and pointed out that “we have a Mythos at home.” Another independent study from Graphistry stated basically the same thing when comparing it with Opus 4.8 and GPT-5.5. Not only that, it achieved excellent results at a fraction of the price: one-sixth of what it cost to run tests with Claude Opus 4.8, for example. Axios revealed little cited a cybersecurity researcher who explained that GLM-5.2 is capable of chaining exploits “in the same way that an elite human attacker would.” Chinese mythos before 2027. Jie Tang, CEO of Z.ai, responded to a Twitter thread in which it was pointed out that at this rate, China would have an AI model at the level of Mythos or Fable by the end of 2026. Elon Musk himself intervened saying that in his opinion this Chinese model with such performance would arrive in the first quarter of 2027. Jie Tang was forceful and replied to Musk saying “it won’t take that long.” And we also have Sakana Fugu. These days we also learned the news that Sakana AI, a Japanese AI startup, had launched Fuguan AI model that is actually not so much an AI model as it is a router or orchestrator of other models. What it promises It is to perform at the level of the best models in the US, taking advantage of different models, both open and closed. The internal benchmarks are promising, but some independent analysts they explained Although the idea is not bad, its performance and cost are not as striking as the company claims. While the US blocks its models, China advances. The situation is paradoxical, because what China is doing is precisely taking advantage of a unique moment. The US is restricting the deployment of the most advanced AI models from Anthropic and OpenAI to avoid cybersecurity risks. And while that happens, Chinese companies are apparently closing the gap with truly remarkable open models. In Xataka | The prompt engineering fashion is over. Now what is important is loop engineering

Richard Liu, CEO of “Chinese Amazon”, points out the fate of 700,000 employees

by

One of the internal debates that some of the largest technology and logistics companies are having to take on revolves around the automation of your templates: replace human workers with robots that do not sleep, do not get sick nor do they demand salary increases. At the same time, they are faced with the dilemma of leaving a good part of their staff unemployed. Richard Liu, founder and CEO of e-commerce giant JD.com, considered the Amazon of China, believes that replace your employees It is inevitable, but consider that technology will “complement” humans, but human labor will find a new space. The key, according to Liu, is for companies to prepare their staff to fill it. The diagnosis, without euphemisms. Liu assured in the framework of the summit of APEC Economic Leaders (Asia-Pacific Economic Corporation) held in Shenzhen that “In the future, when robots deliver packages, the day will come when delivery people will no longer be needed.” But Liu added that he is not going to leave his employees stranded. “Without a doubt, robots will be the ones who deliver the packages. But I really don’t want our 700,000 colleagues to go hungry or lose their jobs,” the manager insisted. The CEO of JD.com no longer proposes a hypothetical replacement of employees with robots, but rather takes it for granted. That is, the question is no longer whether it will happen. What large companies have on the table is when this change will occur, and what is done in the meantime. Amazon was already raising a similar issue with the replacement of 600,000 warehouse employees with robots. JD.com bill more than 150,000 million dollars a year and has more than 900,000 employees. That its CEO speaks in these terms about replacing more than two-thirds of its staff is a very serious matter. The Nirvana plan: 120 schools for 700,000 people. However, JD.com’s approach does not stop at drawing a future of labor collapse, but rather assumes that the new situation will require human labor in other tasks. As I collected Financial Timesthe company has signed contracts with 120 centers education throughout China. Its objective is to train current delivery drivers in robot repair and maintenance tasks in a training program called Plan Nirvana. The idea is that those who today deliver packages on the street end up working in offices programming and maintaining the robots that have replaced them. Liu spoke of “white collar employees” as a destiny for those who are today workers. That means training them as robot technicians, AI trainers and maintenance personnel. The great challenge for JD.com is the scale of converting 700,000 delivery workers into specialized technicians. China: the ground that can sink. Liu’s announcement comes just as a report estimated that China will reach 320 million workers of the “gig economy”. Five years ago there were 200 million. That figure represents about 40% of all urban employment. They are delivery drivers, app drivers or factory workers. People with little economic margin to face a long or uncertain transition that replaces them with robots. However, China seems willing to lead this industrial transformation at all costs and has put robotics at the center of his five year plan approved in March. Xi Jinping’s goal is to make robots the engine of Chinese growth. The government steps on the accelerator of automation and at the same time tries do not overwhelm the most vulnerable with its progress. JD.com, like Amazon, is already doing it. The Chinese trading giant, like its western counterpartalready operates warehouses without staffdelivery drones and autonomous vans in China. At Shenzhen airport, delivery robots They already bring meals at boarding gates, and others they travel by subway to resupply stores. The technology that Liu claims will replace his delivery drivers is already in the testing phase within his operations. Amazon now exceeds one million of robots in its logistics centers and could stop hiring more than 600,000 people until 2033. However, what sets Liu apart is the directness of his speech, which removes some of the uncertainty (and rejection) that are causing this entire process of automation of the labor market among employees. In Xataka | We believed that AI was going to retire an entire generation of workers early. The opposite is happening Image | World Economic Forum, VX Logistics

We believed that no Chinese AI model would soon come close to Fable 5 or GPT-5.5. Then GLM-5.2 arrived

by

A few days ago, the Chinese startup Zhipu AI (Z.ai) announced the launch of its new open AI model, GLM-5.2. It did so boasting amazing features that brought it very close to the best closed models from OpenAI and Anthropic, something that seemed impossible. Well, the more analysis is carried out on the model, the better off it is. We may be at the beginning of something very important. A change of trend. GLM 5.2. The Chinese startup Z.ai has been releasing different versions of its GLM AI model for a long time, but the latest one is undoubtedly the most surprising because its performance is especially promising. It has 744,000 million parameters (744B), of which 40,000 are those that remain active. We are looking at a model with a context window of one million tokens and a new architecture called IndexShare/IndexCache. Better than GPT-5.5, very close to Opus 4.8. The startup showed how the performance of GLM-5.2 is extraordinary in programming tasks. In the FrontierSWE test, the most demanding of those currently available, GLM-5.2 outperformed GPT-5.5 and only Opus 4.8 was superior by a very small margin. The same happened with other tests such as PostTrainBench or SWE-Marathon, which, for example, evaluates the behavior of the model in very long autonomous programming sessions. Source: Z.ai. In many other tests the photo was identical: the model has made a spectacular leap since version 5.1, and is in many tests almost as good (or better) than the best from OpenAI, Anthropic or Google. But it’s not just them who say it.. Artificial Analysis, a reputable independent firm that maintains an updated ranking of the performance of the new AI models that are arriving on the market, confirms the data of Z.ai itself. In his tests he indicates how the “intelligence index” of GLM-5.2 is now 51 points. It is only surpassed by GPT-5.5 (55), Claude Opus 4.8 (56) and Claude Fable 5 (60). Source: Artificial Analysis. This Chinese open model leaves behind the new Gemini 3.5 Flash, but also Chinese competitors such as Qwen 3.7 Max, MiniMax-M3 or DeepSeek V4, among others. The jump in quality from GLM-5.1 is, we insist, outstanding, much greater than what, at least according to this index, was seen from Opus 4.8 to Fable 5. The jump in performance is spectacular, although it is true that the comparative price to solve the tasks proposed in the benchmark rises significantly. Source: Artificial Analysis. But it’s not perfect. The Artificial Analysis report, however, shows that although GLM-5.2 is very strong in areas such as programming, it is weak in others. For example, it is far from being as reliable as Fable 5, GPT-5.5, Claude 4.8 or Gemini 3.1 Pro in terms of correct answers, which is also lower in proportion to that of its competitors. However, his hallucinations have significantly reduced. And it’s much (much) cheaper. But in addition to being fantastic in many areas, it is much cheaper than its competitors. Maintains the price per million input/output tokens of its predecessor ($1.4/4.4), while that of GPT-5.5 It’s 5/30 dollars and that of Opus 4.8 10/50 dollars. It is true that it consumes many more tokens than GPT-5.5 (very efficient) or Claude Opus 4.8, but even with that its final cost is much lower. My tests with GLM-5.2 programming. I’ve been a Z.ai subscriber for months now because they offered an annual subscription at the end of 2025 at a really low price. This has allowed me to test GLM-5.2 for a few hours and although I cannot draw definitive conclusions, it does seem clear that there is a leap in quality in terms of its ability to program. I asked him to review a personal code project and he identified several security flaws and possible improvements in great detail. Chatting with GLM5-2. In conversational mode the behavior is much more difficult to evaluate: I have been interacting with the model and asking it questions, and although it is better than GLM 5.1 many times, other times it is not so much and I would say that in terms of creativity to write the frontier models of Google, OpenAI and especially Anthropic they are still quite superior. You can try it on their websiteand there you will see something else: it takes significantly longer to respond than other chatbots, because its reasoning phase is longer. Take more time to answer questions. Benchmarks are one thing, experience is another.. In the absence of testing it (much) more, of course the impression is that the model has improved significantly compared to a GLM-5.1 that had lagged behind its Chinese competitors (not to mention the current Claude Opus 4.8 or GPT-5.5). On platforms like Reddit opinions are dividedbut many consider it a fantastic option to run locally… if you have a very, very powerful machine with at least 256 GB of unified memory (Mac Studio). And one thing seems clear: when using it as an AI model for programming, comes surprisingly close to Claude Opus 4.8. In Xataka | Chinese technology companies entered the AI ​​race with cheaper models than the rest. That’s starting to end

The new Chinese gem of semiconductors is called Enflame. This is the new member of “the four chip dragons” of China

by

The name Enflame may not ring a bell yet. But it is very likely that in the coming months it will end up giving us a lot to talk about. And this Chinese AI chip company just got the go-ahead to go public on the STAR market in Shanghai, the preferred market for the country’s large technology companies. After this, we see how the scheme of large chip manufacturers begins to take shape. Enflame enters the select group of the four big technology companies that are dedicated to AI chipsand that are already listed or are about to do so on the public markets. Who is Enflame and where does it come from? The company was founded in Shanghai in 2018 by Zhao Lidong, an engineer who came from AMD, where he led the development of high-performance processors at the American company’s R&D center. Together with his co-founder Zhang Yalin, Zhao set out to replicate that knowledge in Chinese territory and build a domestic alternative to Nvidia. In seven years has developed five AI chips distributed across four generations of architecture, and has built a catalog that includes processors, accelerator cards, computing clusters and software platforms. Its most recent chip, the L600 module, has passed silicon verification testing, although it has not yet entered large-scale commercial production. Why this IPO matters. Enflame plans to raise up to 6 billion yuan (about 888 million dollars) selling between 10% and 15% of its shares. The money, as could not be otherwise in these times, will be used to accelerate the development of its next generation of AI chips in the cloud and build the software that surrounds them. However, the operation also has a certain symbolic character, since it is the fourth and final addition to the group known as the “four little dragons” of Chinese chips. The other three (Moore Threads, Biren Technology and MetaX) have already debuted on the STAR market, and have been received enthusiastically by investors. In fact, Moore Threads, nicknamed “the Chinese Nvidia”, rose 425% on its first day of trading in December of last year, according to Bloomberg. Restrictions. The reason China is betting so big on these manufacturers is that the United States has been applying restrictions on chip exports for years advanced towards the Asian giant. Nvidia’s most powerful models are blocked, which has created a real shortage in the Chinese market and a strategic urgency to develop its own alternatives. Beijing has responded with public moneyincluding a relaxation of STAR board rules to allow loss-making companies to list, and a $295 billion plan to build data centers that do not depend on American chips. In this framework, Enflame and its groupmates become part of an infrastructure of technological sovereignty. What does it look like? Tencent. Enflame’s greatest asset is also its greatest vulnerability. Tencent owns about 20% of the company and in 2025 it represented 84% of its income, compared to 38% the previous year. That is, almost everything that Enflame sells is bought by Tencent. The Chinese tech giant uses its chips to power large-scale data centers, recommendation systems, chatbots and generative AI infrastructure. The company itself acknowledged in its IPO prospectus that “Tencent’s demand has far exceeded its supply capacity.” That’s good in the short term, as it guarantees income. But how they point out In The Next Web, a chip maker that relies on a single customer for the majority of its sales ends up being exposed if that customer changes priorities. The numbers. Enflame is growing at breakneck speed, as revenues have multiplied a compound rate greater than 80% between 2023 and 2025, but still in losses. Net losses were reduced to 1.2 billion yuan in 2025, compared to 1.5 billion the previous year, and the company plans to close the first half of 2026 with losses of about 600 million yuan. For the same period, it expects its revenue to grow more than three times compared to the previous year, reaching between 10.6 billion and 11.5 billion yuan. On the other hand, investment in R&D has exceeded 100% of sales over the last three years, which says a lot about the phase the company is in (still building, not harvesting). Before the IPO, the Hurun Index valued the company at around $2.8 billion. Where Enflame fits in. Not all dragons are the same. Within China, Enflame competes in a market where Huawei and Cambricon They continue to be the benchmarks in the sector and are already profitable. Enflame, Moore Threads, Biren and Iluvatar CoreX make up a second, younger layer that is trying to break through. Technically, Enflame has opted for application-specific integrated circuits (ASICs), a more specialized architecture, rather than the general-purpose GPUs used by Moore Threads or Biren. Xu Dawei, of Jintong Private Fund Management in Beijing, points out Bloomberg that Enflame “benefits from solid comparatives,” given that its Chinese competitors are already listed on the stock market with valuations well above what their revenues would justify. Companies like ByteDance are actively looking for domestic alternatives to Nvidiaand second-tier manufacturers, including Enflame, are on the radar. Cover image | Enflame In Xataka | TSMC is on the ropes and its biggest problem is not competition: it is water

An unexpected salvation for the end user emerges from the memory market debacle: Chinese chips

by

The DRAM memory industry is facing a profound structural transformation. Until October 2025 the price of memory chips evolved in a relatively stable way, but from that moment on began a dizzying climb which still continues. In fact, the consultant TrendForce expects the price of conventional DRAM to rise between 58% and 63% quarter-on-quarter before the expiration of the second quarter of 2026. And the artificial intelligence (AI) is behind all this. The three largest chip manufacturers of memory on the planet, the South Korean companies SK Hynix and Samsung Electronics, and the American Micron Technology, They have reallocated about 70% of its production lines to high-bandwidth memories (HBM) to satisfy the currently insatiable demand of data centers specialized in AI. The consequences of this movement did not take long to appear: standard DDR4 and DDR5 memories and their derivatives, which are the most used in the consumer segment, immediately began to become scarce. And its price skyrocketed. In fact, according to the consulting firm GartnerRAM has gone from representing 16% of the total cost of a laptop to 23%. And it is possible that this escalation will continue to develop in the coming months. However, users can cling to the greatest stabilizing agent in the memory market today: Chinese manufacturers. This is the great opportunity for YMTC and CXMT Yangtze Memory Technologies Co. (YMTC) is one of the largest NAND chip manufacturers in China. Its global market share is approximately 13%making it one of the main competitors of Samsung, SK Hynix, Micron, Kioxia or SanDisk. Its weight in the Chinese market is very great, especially because US sanctions They prevent American and South Korean memory manufacturers from selling their most sophisticated integrated circuits to their Chinese customers. On the other hand, Changxin Memory Technologies (CXMT) is one of the Chinese companies specialized in the production of memory chips, and, like other companies in the country led by Xi Jinping, it has chosen to compete in this very attractive market by deploying a very aggressive pricing policy. CXMT in particular has increased its DRAM chip production capacity almost five times during the last four years, which has allowed it to increase its global market share until reaching a very worthy 7.6%. CXMT has chosen to compete in this very attractive market by deploying a very aggressive pricing policy. While large foreign manufacturers maximize their margins thanks to data centers and the rise of AI, Chinese manufacturers prioritize sourcing from local companies. This scenario allows the supply and prices of memory and NAND chips in China to remain relatively stable, remaining outside the strong premiums charged by the big three (Samsung, Micron and SK Hynix). This is the context in which the Chinese memory module brands Gloway and KingBank have recently announced new DDR5 modules that stand out for using SDRAM memory chips made in China. With a standard configuration of eight chips per module, these companies can produce 24 GB modules and group them into kits of two or four modules to achieve capacities of 48 GB or 96 GB, respectively. Chinese memory chips, particularly those from CXMT, have already begun to spread beyond China’s borders. Corsair has already integrated them into some kits of its Vengeance line, while HP and Dell have begun the process of homologating modules with CXMT chips for their products. This is good news for users, there is no doubt. Even so, we still don’t know if the use of CXMT DRAM will become widespread in response to AI-induced shortages. The market demands new players, wherever they come from, and if YMTC and CXMT are able to fill the gaps left by Samsung, Micron and SK Hynix, they are welcome. Image | Intel More information | Tom’s Hardware In Xataka | China needs to develop a new type of chips immune to US sanctions. And your scientists have just achieved it

A Chinese billionaire bought the most expensive house in London. What happened next is a real estate horror movie

by

In 2021, the Evergrande collapse It erased more than $300 billion in liabilities and triggered a real estate crisis that shook all of China. Among his most extravagant assets was a record mansion in London that today remains trapped in lawsuits, divorces and frozen accounts: a financial ruin turned into an empty monument. The perfect purchase that went wrong. Year 2020, a Chinese billionaire buys for 210 million pounds number 2-8A Rutland Gate, then the most expensive home ever sold in the United Kingdom. On paper it was the definitive investment: a palace with 45 rooms, four elevators, an indoor pool, 24 marble bathrooms and privileged views of Hyde Park. But what seemed like a prestige move ended up leading to a chain of misfortunes so strange that it seems written like a thriller. Since then no one has lived inside, the real owner was caught in a financial collapse and the building became an empty shell with only one “tenant”. Palace with a cursed past. The history of the building was already coming loaded with symbolism. For decades it was the London palace of Rafik Haririwho transformed it by joining together several Victorian houses and decorating it with almost obscene luxury, from gold-plated trash cans to bathrooms encrusted with semi-precious stones. Hariri was assassinated in Beirut in 2005 and, after passing through the hands of the Saudi royal family, the interior was auctioned piece by piece in 2015. That left the mansion empty, as if it had been dismantled before its next owner arrived. The Evergrande turn. The official buyer of 2020 appeared to be the Hong Kong tycoon Cheung Chung-kiubut later was discovered that the real owner was Hui Ka Yanfounder from Evergrande and for years the richest man in China. And there the descent began. Just a year later, Evergrande began with non-payment of debtsbecame a symbol of the Chinese real estate collapse and ended up collapsing in 2024. Hui ended up declaring guilty of fraud and other financial crimes, while the mansion was trapped in a legal tangle: registered in the name of his ex-wife, with frozen assets and no possibility of sale. The most expensive house in the United Kingdom lost in limbo and a symbol of the real estate gap. Fernstedt at the entrance to the house The empty house and the Swede on the porch. And it is at this point in history where the image appears that changes everything to this day. While inside the mansion there are dozens of empty rooms and millions of pounds tied up, outside, on the same porch, lives Anders Fernstedta homeless Swede who has been living at the entrance for three years. Your “camp” It is made of umbrellas, flowers, broken bicycles and stuffed animals. The paradox could not be starker, because he sleeps inches from one of the most crazy expensive shelters in Europe, but separated by a door that never opens. Ironically, the only stable inhabitant of the house does not have access to it. The fall of Anders. As to the history of man Swedish, is almost as chaotic as the architecture on which it rests. Andres was a technology journalist, then he worked in horticulture, collaborated with people from the Silicon Valley environment and even worked for The Economist as a freelance fact-checker. However, a chain of failed jobs, evictions, attacks and personal losses dragged him onto the streets. He ended up landing in front of the palace purely by chance: He was looking for a covered shelter and found an empty porch. He has since converted that space into a kind of makeshift garden and permanent bedroom. The symbol of a broken city. It had an extensive Guardian report that the story of Rutland Gate sums up a huge contradiction in London. While more than 300,000 homes remain empty in England and hundreds of thousands are waiting for a house, one of the most luxurious properties in the country has been closed for years because it is, in reality, a frozen financial asset. From that perspective, it is no longer a home, it is a figure trapped in offshore companies, lawsuits and bankruptcies. And in front of that door, every night, a man sleeps what represents just the other end of the system: someone with nothing, living in the shadow of a palace that no one can use. Image | Gareth E. Kegg In Xataka | Now that the Pope is in Spain, he should visit this surgeon’s castle. Inside is the smallest church on the planet In Xataka | In 1972 Italy wanted to put an entire city in a one kilometer building. Half a century later he is still paying the consequences

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.