pretty Archives - usatoday24

DeepSeek is good, pretty and very cheap. And above all, the weapon to create a Chinese hardware industry independent of Nvidia

May 25, 2026 by usatoday24

The arrival of DeepSeek-V4-Pro It hasn’t caused that much of a stir. like the one caused by DeepSeek R1 a year and a half ago, but we may be facing an even more important model. If that version revealed to the world that China was advancing spectacularly in this race, this other one is beginning to allow us to glimpse something else more interesting. What most people see is a very decent model and above all “low priced”. Which hide the company It’s another more important thing: achieve independence from Nvidia and US hardware. what has happened. Last Friday, those responsible for DeepSeek announced something surprising: their promotional offer with a 75% price cut to use their DeepSeek-V4-Pro model will be maintained permanently. That makes this model offer very decent features (but not exceptional) for a really low price: 1M entry tokens 1M tokens output DeepSeek-V4-Pro 0.435 0.87 GPT-5.5 5 30 Opus 4.7 5 25 Gemini 3.5 Flash 1.5 9 Good, pretty and very cheap. It is true that the performance of DeepSeek-V4-Pro is inferior to that of rival models from OpenAI, Anthropic or Google. Artificial Analysis tests indicate that the DeepSeek model is at a very good level, but it is also much cheaper than its competitors. This is especially relevant for agentic tasks that consume many tokens and that with this model become accessible and very affordable. According to Artificial Analysis, DeepSeek is close to the performance of the best models in the industry, and although it is slower in its responses, it is also much cheaper than the frontier models from OpenAI, Anthropic or Google. A different strategy. How is this company going to make money? It does not have subscription plans like its local competition (GLM, Kimi) or the western one (ChatGPT Plus, Claude Pro). It also does not have voice or image models. It does not have an AI agent for programming that competes with Claude Code. It publishes the open weights of its models and shares its technical innovations with the industry (and with its competitors). For those who closely follow the company and these decisions, the strategy is clear. DeepSeek’s goal is not to win the AI model race. Their goal is to build a Chinese AI hardware industry that doesn’t depend on Nvidia or TSMC… and get paid their share in that process. Hardware independence. China has a structural problem in this AI race: sanctions and vetoes imposed by the US make you unable to access the most advanced chips nor to ASML UVE photolithography. And since China cannot currently compete in terms of computing power, what its companies are doing is ensuring that their AI models need less computing power to achieve similar results. Efficient architectures. The Mixture of Experts (MoE) and Multi-head Latent Attention (MLA) architectures are two key weapons in this strategy. The first already existed but was adapted by DeepSeek for their model: with it only part of the total parameters of the model are activated to answer the query without losing precision. What MLA does is compress the attention information (the so-called KV Cache) with which the model maintains the context of a conversation, reducing it by 90%. Both techniques allow us to reduce the need to use high-speed HBM memories, something that is also striking in order to reveal DeepSeek’s probable strategy. The importance of KV Cache. As the GDP analyst explains in Xthat use of MLA allows that for one million tokens, DeepSeek-V4-Pro only needs 5.48 GB of HBM memory. Competitors like Zhipo AI, which develops GLM 5, need 60 GB for the same, while Alibaba’s Qwen 3 needs 89 GB. This advantage allows DeepSeek to offer much lower prices to obtain performances similar to those of its competition, but it also means that DeepSeek models can run on Chinese memory chips that cannot compete in speed with HBM modules. Goodbye HBM, hello NAND and SSD. These innovations open the door to the use of NAND memories and even SSD drives to process this data, and there YMTC enters the scenea Chinese Flash memory manufacturer that is slowly becoming a global giant. Also CXMTwhich manufactures DRAM memoriesbecomes an alternative here and the reason is equally interesting: DeepSeek introduced a memory search module in LLMs called Engram which is also intended to avoid excessive dependence on HBM memories. How to bypass the CUDA monopoly. Nvidia continues to have a fundamental element in CUDA to maintain its market dominance, but here DeepSeek too has proposed an alternative. Is called Tile Kernels and these are software cores created with TileLang (a variant of Python for this field) that allow governing advanced AI chips (GPUs). Huawei as an invisible ally. Those responsible for Huawei recently indicated that its new Ascend AI supernodes fully support DeepSeek v4 models. Precisely this provides another fundamental advantage to the company, which thus avoids (at least in part) total dependence on the use of Nvidia chips and prepares to further strengthen Huawei’s relevance in a market in which until recently Jensen Huang’s company was queen and mistress. Open models to attract the hardware industry. US companies continue to maintain their closed and proprietary models, but DeepSeek is one of the many Chinese startups that publish them with open weights. With this, what she and the others intend to do is not only attract AI developers and users, but also create a hardware ecosystem that adopts these architectures. DeepSeek invites its rivals to use techniques such as MoE or MLA precisely so that all these advances become a de facto standard and hardware manufacturers also adopt them and integrate them in an optimized way into their designs. A round of 10,000 million to advance. The company is also preparing a financing round in which they intend to raise 10,000 million dollars and with which they would achieve a valuation of between 45,000 and 50,000 million dollars. Still far from the mammoth valuations of OpenAI or Anthropic (already close to a billion dollars) but certainly … Read more

There is a battle to have the AI model that programs best. And a good, pretty and very cheap rival has appeared in it: Cursor

May 19, 2026 by usatoday24

Cursor has introduced Composer 2.5a generative AI model specifically intended for one thing: programming well. How good? Well, according to this startup, it does it as well as the best models of the moment, Claude Opus 4.7 and GPT 5.5, but it also does it for a lower cost. The challenge is striking not only because of what it means for Cursor, but because of how they have created that model: it turns out that it is based on a Chinese AI model. AI models specialized in one thing. While OpenAI and Anthropic try to develop general-purpose models—they do a lot of things really well— Cursor you have decided to focus on a specific task. The AI startup has created an AI model specialized in programming, and has done so by arguing that a billion parameters are not necessary to compete with the best. Devoting yourself to a single thing allows you to not only gain efficiency, but also costs. This is not a decathlete, but a specialist in the 200 m event, so to speak. As good as GPT-5.5 or Claude Opus 4.7? That’s what they say in Cursor, because according to their tests with several specific programming benchmarks, the performance is on par with those two models that today are the great references both in programming and in other areas. And much cheaper. These results are also especially interesting when we add the cost factor. The average cost per task in the CursorBench 3.1 benchmark showed that Composer 2.5 managed to solve almost 65% of all tests for a cost of just $0.3. Opus 4.7 max and GPT-5.5 xhigh managed to reach that 65%, but at much higher costs: just over 4 dollars in the case of GPT, and 11 dollars in the case of Opus. The difference is abysmal. He API access price demonstrates the differences: 0.5 dollars per million input tokens 2.5 dollars per million output tokens, when Claude Opus 4.7 is 5/25 and that of GPT-5.5 is 5/30 respectively. Textual feedback. Unlike models that only learn from the final result, Composer 2.5 has been trained with a reinforcement learning technique (Reinforcement Learning) that allows us to offer clues about what is happening if errors are being made. This allows the model to recalibrate and act as a transparent teacher. One that also corrects word by word as it solves the exercise, not just when seeing the final result. 85% of the training budget has been dedicated exclusively to reinforcement learning, calibrating the model not for chat, but to execute code refactorings or fix bugs in real time. A model “born” in China. Those responsible for Cursor themselves have explained that Composer 2.5—like its predecessor, Composer 2launched at the end of March—is a model derived from Kimi K2.5, the AI model of the Chinese startup Moonshot. Although that is the basis, already in Composer 2 the training and post-training tasks manage to improve the behavior in a very notable way in programming benchmarks and also in others such as Terminal Bench that evaluate the agentic behavior of these models. Cursor gets older. This startup became famous for creating a programming AI agent that was a pioneer in that fever we live for vibecoding. The user experience is no longer that of programming, as in traditional IDEs (Integrated Development Environments), but rather that of directing the machine to program it for you. Composer 2.5 doesn’t just program: it understands the structure and relationships between files, and turns Cursor into a much more competitive AI company, because it no longer depends on being able to work with Anthropic or OpenAI models, for example. Having both the AI agent and the model processing everything makes it a much more competitive solution. Elon Musk has Cursor in his sights. Cursor’s good performance has led to growing interest in buying this company even before it becomes too big. Elon Musk knows this well and Grok, xAI’s model, is not so popular in the programming field. In April we learned that SpaceX had reached an agreement that gives you the option to buy Cursor for 60,000 million dollars. It would be a promising deal for both, because Composer 2.5 has already used Colossus’ infrastructure to train, and xAI could thus try to gain market share in the juicy enterprise sector. In Xataka | Elon Musk knows that TSMC is overwhelmed: Terafab is his idea to completely change the global chip industry

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

April 8, 2026 by usatoday24

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

There is an island that has been prohibiting cars for more than a century and continues to use horses. And it’s doing pretty good

July 5, 2025 by usatoday24

Before entering the subject, let’s make a game: Open Google Maps, Type “Mackinac Island” And let the search engine transfer you to a small island in Lake Huron, in Michigan, USA. Then approximate and handle Street View to take a virtual walk through its streets. In their wide avenues you will see people walking, people by bicycle and people mounted in carriages thrown by horses, but what you will not find are cars. Very lucky maybe CACES some (few) of those used promptly to provide certain services. After all Mackinac is known worldwide Therefore: have banished motor vehicles and stay, in the middle of 2025, such as the reign of horses. In a Michigan place … Mackinac is a fantastic example of how history is full of ironies. The island known inside and outside the US for its aversion to cars is located in the middle Ford, General Motors either Chrysler. In fact Detroit, the “Motor City” It is located just 400 k, in a straight line. However, despite this influence of the industry 127 years ago the island authorities made a peculiar decision: they prohibited the use of combustion vehicles. Petarders no, thanks. The veto was promulgated the July 6, 1898after the islanders who dedicated themselves to working with calese alerted of the “dangers” and discomfort that the new “carriages without horses.” The legend He says that the trigger (never better) was the firecracker of a vehicle that in 1898 frightened a group of horses. It is also not far -fetched to think that the chauferes They moved to shield your business in the face of engines. A prohibition in DNA. The fact is that the rule set. After a few years It extended To the rest of the island, just 3.8 km2, and with the passing of the decades it became one of Mackinac’s great hallmarks. Little served to mark like Oldsmobile Or Ford became stronger and more only a few kilometers from there, the small island of Lake Huron remained an impregnable redoubt for the thriving automotive industry and thus has continued to be during the twentieth and twenty -first centuries, for pride of local authorities. On their official website they remember that the M-185the road that surrounds the island, is the only US state road in which the use of motor vehicles is not allowed. “Of the more than six million kilometers of public streets in the US, there is a stretch of 13.2 kilometers in Mackinac that stands out for its uniqueness,” They need The authorities. “It could be decided that it is literally one among one million.” There is no car? No. And yes. The authorities do not allow people to use cars as they would do any other part of the US, but that does not mean that there are certain exceptions. Mlive remember That the island has emergency vehicles, police cars, an ambulance released in 2021 and fire trucks. The State Park also has vehicles, although only uses them out of season high and preferably at first or last hour of the day. The island has also turned a blind eye in certain cases. For example, the US secret service put a car in 1975, during a visit by President Gerald R. Ford with his wife. The vehicle was used by agents. Ford opted for a carriage pulled by horses. Another exceptional case was the filming in 1979 of ‘Somowhere in Time’a film starring Chistopher Reeve and Jane Seymour that was shot in Mackinac and had a special permission to use vehicles. And the rest of the time? Simple. People walk, move by bike or ride in carriages thrown by horses, one of the great hallmarks of the island. It is believed that the horses arrived in Mackinac around 1780 by the British, who used them to lift the Strong Michilimackinacand the caleses were popular mostly in the nineteenth century, when the island became popular as a resting place. The first license for a carriage was issued in 1869. The island also has a ferry that allows its 600 inhabitants Move more easily from neighbors Mackinaw City or St. Ignace. “Without horses, this place would not be what it is. It allows you Michigan’s Upper Peninsula. A destination with history. With the passing of the decades Mackinac has achieved more than becoming a small redoubt safe from the traffic and contamination of cars. It has also become a popular destination, especially during The summer months. There, in addition to their horse carriages, coasts and landscape The Anishnaabe culturea group of indigenous peoples from the region of the Great Lakes of North America. Images | Dan Gaken (Flickr), Greg Marks (Flickr), Kate Ter Har (Flickr) and Poissantfamily (Flickr) In Xataka | In the US there is a “colonized” city by the Basques. And it has its own Ikastola, Frontón and Ikurriñas on the street

Log In

Sign In

Forgot password?

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections