models

running AI models

by

It was “sung” that Apple would present its new Apple M5 sooner rather than later, and we already have it with us. The chip represents a notable qualitative leap judging by its specification sheet, and there are improvements in all sections. The surprise is that where a real change is seen is in its GPU, which is now much more powerful and is clearly prepared to be able to work with AI models in a much more striking way. Apple M5. Apple’s new SoC makes use of third-generation 3nm photolithography. There are therefore no major changes in the manufacturing process, which is surely more efficient and reliable. Even so, the changes in the chip are notable everywhere. To start, the CPU has six high-efficiency cores and four high-performance cores. According to Apple, this CPU offers 15% more multicore performance than the M4. Working with local AI models thanks to apps like LM Studio is going to be much more feasible thanks to these Apple M5s. A GPU with a lot of potential for AI. If there is a standout element in this SoC, it is the GPU, which in the M5 has 10 cores, each of them including a Neural Accelerator. According to Apple, this allows AI workloads to run “dramatically faster”, and promises performance of more than four times that achieved with the Apple M4. There are also improvements in the bandwidth of the unified memory, which is now 30% higher and reaches 153 GB/s, something especially crucial for using AI models locally. These chips allow for configurations with up to 32 GB of unified memory: the figure is not particularly high, but we will undoubtedly see an M5 Pro and an M5 Max with much more room for maneuver in this section. And also to play. There is a lot of good news in this GPU for those who plan to take advantage of it to work with local AI models, but there are also improvements for video games. Graphics performance is up to 45% higher than the M4 GPUs, and we also have third-generation ray tracing technology. The neural motor is strengthened. The new Neural Engine has 16 cores and combines, according to Apple, a mix of efficiency and performance. This means that, for example, the new Vision Pro, which also includes this chip, can transform 2D photos into spatial images in the Photos application easily and quickly. Prepared for the future of Apple Intelligence. Apple Intelligence may not be a notable artificial intelligence platform today, but it is clear that Apple is preparing its devices for a scenario in which being able to run AI models locally (and privately) is the norm. The recent iPhone 17 also made a striking leap in these capabilities with the A19/A19 Pro, and now we see the same movement with these M5s that put all the focus on AI. Waiting for the M5 Pro/Max/Ultra?. These chips can already be found in the first products that use them: the new 14-inch MacBook Pro, the iPad Pro and the Vision Pro (2025). It is to be expected that Apple will soon present more powerful versions of the recently launched M5, because that is what it has done in previous generations. Last year he launched the Apple M4 in May and M4 Pro/Max at the end of October. So five months passed. In the previous generation the M3, M3 Pro and M3 Max were launched in October 2023, and the M3 Ultra appeared much later, in March 2025. The M4s have not seen (at least, for now) an Ultra version, but Apple may end up doing something similar to what it did with the M3s. If they follow the previous scheme, the M5 Pro/Max would be announced in February or early March, and more or less around those dates a hypothetical M4 Ultra could appear. Rumors suggest, however, that these chips would arrive later, in mid-2026. Future teams. Meanwhile, yes, we can now enjoy the first devices that have these chips. It is also expected that Apple will offer new devices with the M5 chip in the middle of next year: that is when the Mac mini with M5, the iMac with M5 and the Mac Studio M5 and on the other hand the MacBook Pro with M5 Pro/Max are expected to arrive. In Xataka | Apple has become a boring company. We wonder who will inherit his throne: Crossover 1×24

These televisions are today (October 13) at outlet prices: models from 169 euros

by

The television is one of those essential devices, for the vast majority of users, in any home. If you are thinking of renovating yours or want to buy one for a bedroom or even the kitchen. Today, we have found some bargains on TVs that may interest you. Daewoo 43DM56UV by 169 euros: 43-inch LED and with Android TV operating system. Toshiba 43UV3463DG by 229 euros: 43-inch Direct LED and 4K Ultra HD resolution. Philips 50PUS8010 by 329.99 euros: 50 inches and with Ambilight. Hisense 55E7NQ Pro by 499 euros: 55-inch QLED with 144 Hz. Samsung TQ75Q64DAUXXC by 949 euros: 75-inch QLED and with Tizen. Daewoo 43DM62UA We begin our compilation with one of the cheapest televisions that you can find on the Carrefour website. It’s about this Daewoo 43DM56UVwhich presents a very good quality-price ratio. Its usual price is 219 euros, but now it is reduced to 169 euros. This cheap TV has a 43 inch LED panel and with 4K Ultra HD resolution. Its two speakers offer a power of 16 W and it comes with Android TV operating system. It integrates WiFi, Bluetooth and Ethernet and comes with three ports HDMI and two USB 2.0 ports. The price could vary. We earn commission from these links Toshiba 43UV3463DG Another of the cheap TVs that you can buy at Powerplanet is this one Toshiba 43UV3463DG. Usually, it costs 299 euros, but now it has a 34% discount applied, which means that you can purchase it for 229 euros in these moments. This smart TV mounts a panel 43 inch Direct LED with 4K Ultra HD resolution. Its speakers offer an RMS power of 16 W and it comes with VIDAA operating system. It is compatible with HDR10 and Dolby Vision and is also compatible with Google Assistant and Alexa. Toshiba DLED 43″ 43UV3463DG UltraHD 4K VIDAA The price could vary. We earn commission from these links Philips 50PUS8010 If you want to have a TV with Ambilight at home, this model on offer at PcComponentes will interest you. Now, you can get the Philips 50PUS8010a 50-inch model that usually costs around 350 euros, but now you can take it for 329.99 euros. This TV mounts a 50-inch LED panel with 4K Ultra HD resolution. It works under Titan OS operating system and has the lighting system Ambilight. Its two speakers offer a power of 20 W and are compatible with Dolby Atmos and it comes with three HDMI 2.1 ports, two USB ports and is compatible with Alexa and Google Assistant. Philips Ambilight 50PUS8010 4K LED Smart TV The price could vary. We earn commission from these links Hisense 55E7NQ Pro Another of the bargains on televisions that are worth it at PcComponentes is this one. Now, you can take this smart TV Hisense 55E7NQ Pro. With a 28% discount. Its usual price is 569 euros, but now you can get it for 499 euros. It is a TV with a 55-inch QLED panel with 4K Ultra HD resolution and 178º viewing angles and compatible with HDR10+. Its speakers offer a power of 40 W and are compatible with Dolby Atmos and DTS Virtual:X. It has a refresh rate of 144 Hz and AMD FreeSync Premiummaking it ideal for gaming). Finally, it can be noted that it works under the VIDAA operating system. Hisense – QLED TV 139 cm (55′) Hisense 55E7Q Pro, UHD 4K, Smart TV. The price could vary. We earn commission from these links Samsung TQ75Q64DAUXXC And if what you are looking for is a large TVthis 75-inch Samsung Q64A is another of the bargains that you can find now at MediaMarkt. Its usual RRP is 1,079 euros, but now it is available for 949 euros. Assemble a panel QLED with a diagonal of 75 inches and 4K UHD resolution. It supports HDR10+ and comes with Filmmaker Mode. Works under the operating system tizen and it has speakers with a power of 20 W. Finally, it is worth mentioning its connectivity section, since it integrates WiFi 5, Bluetooth 5.2, Ethernet, three HDMI, two USB and optical audio output. Samsung QLED Q64D 75″ 4K UHD Smart TV AirSlim Quantum HDR Q-Symphony The price could vary. We earn commission from these links Some of the links in this article are affiliated and may provide a benefit to Xataka. In case of non-availability, offers may vary. Images | Webedia, Daewoo, Toshiba, Philips, Hisense and Samsung In Xataka | Best home theater projectors. Which one to buy and five recommended models from 299 to 18,000 euros In Xataka | Best sound bars in quality price. Which one to buy and seven recommended models from 140 euros

Alibaba has one of the best open source AI models. Your next step: use it in robotics

by

Alibaba has taken another step in its commitment to artificial intelligence by creating an internal team dedicated to roboticswhich will operate from qwenits AI modeling division. The Chinese giant, owner of one of the best open source AI models, now wants the Qwen team to know how to apply its knowledge in robotics, a sector that is beginning to awaken interest, not only in industry, but also with the arrival of projects in the domestic sphere. Who leads the project. Justin Lin, technology manager at Qwen and expert in multimodal models (capable of processing text, sound and images), was the one who has confirmed the creation of this “small team for robotics and embodied AI” through their social networks. Lin has worked on the development of the Qwen models, which are currently among the most popular in open source globally. The vision behind the movement. According to Linmultimodal AI models are evolving into “fundamental agents” capable of performing complex long-term reasoning tasks thanks to reinforcement learning. “They should definitely make the leap from the virtual world to the physical world,” he said. explained the manager, making clear the intention to apply these technologies in tangible devices. Alibaba’s big bet. This announcement is part of Alibaba’s broader strategy in the sector. Last month, the company led a financing round of 140 million dollars at X Square Robot, a Chinese robotics startup. In addition, its CEO Eddie Wu esteem that global investment in AI will reach $4 trillion in the next five years, a figure that reflects the sector’s expectations. Global competition. Alibaba is not alone in this race. Nvidia and SoftBank are also making significant moves in smart robotics. SoftBank just announced the acquisition of ABB’s industrial robots business for $5.4 billion, while Nvidia CEO Jensen Huang has qualified the combination of AI and robotics as a “multi-billion dollar” long-term growth opportunity. China is also the world’s leading power in the robotics sector. And only in 2024, Chinese factories installed nearly 300,000 industrial robotsa figure higher than the rest of the world combined. The Qwen factor. The choice to place this team within Qwen makes all the sense in the world. Seven models of the Qwen series are currently listed in the top 10 Hugging Facewith the multimodal model Qwen3-Omni occupying first place. This strength in AI provides the company with a solid foundation to develop advanced robotic applications based on the journey they already have with Qwen. Cover image | zhang hui and Possessed Photography In Xataka | AI companies have just encountered an unexpected challenge: insurers have started to turn their backs on them

One of the most downloaded apps for iPhone pays for recording calls to train AI models. It is a security disaster

by

The sale of personal data is not a hypothesis, it is an expanding reality. Just look at Spotify: Recently a service appeared that paid those who delivered their profile and their listening summaries to resell them to technology companies. The approach was as simple as disturbing, because it became something as innocent as our musical habits. Neon Repeat the scheme, but transfers it to a much more sensitive land, telephone calls, where intimacy becomes the product. We are talking about an app that decided to convert phone calls into the new digital gold. His proposal is direct: “Speak, record and charge.” It promises users to win “hundreds or even thousands of dollars a year” simply allowing their conversations to transform into training material for artificial intelligence systems. The hook worked. In a matter of days he went from irrelevance to place Within the top three positions In the Social Networks category in the United States App Store. How neon works. The neon mechanism is designed for each call to translate into money. It promises to pay 30 cents per minute when two users of the app talked to each other, 15 cents if the call is with someone external and establishes a stop of 30 dollars daily. To this adds a referral system that offers 30 dollars for each new user. The recording, According to your policyalways affect the sender and, when both used neon, to both parties. Conditions of use. Beyond payments, the true neon reach is in its Terms of service. There the users give the company a “world, exclusive, irrevocable and transferable” license on their recordings. This permit includes rights to sell, modify, create derived works and distribute the audio in any format, present or future. To this is added a section of functions in beta, without guarantees or responsibility in case of failures. The amplitude of that assignment makes it difficult to foresee how far the use of the recordings can go. Where is available and how popular it is. Neon’s initial success was as fast as unexpected. At the time of writing this article, it is number 2 of the most downloaded social applications in the United States App Store. The application, however, seems restricted to that market: in tests carried out from Spain is not among those available or allows its download. The security failure. The story took an unexpected turn when a technical analysis revealed that Neon did not protect the information of its own users. As Techcrunch discoveredjust create an account and review network traffic with a tool like Burp Suite to access others. Shortly after the notice, the founder closed the servers and sent an email announcing a pause ‘for security’, not to mention the filtration. What was exposed was especially delicate: Telephone numbers associated with accounts Public links to audio recordings Complete call transcripts Metadata with duration, date and payments obtained Telephone numbers, recordings and transcripts are not accessible is not a minor failure. With this data, private conversations could be rebuilt and associated with specific people. The risks range from attempts to impersonate identity to the creation of synthetic voices. What Neon says in front of what we know, Neon defends that their processes protect users: anonymity of conversations, elimination of personal information and sale only to reviewed companies. However, the ruling showed that these systems are not infallible. The official communication after temporary closure spoke of “adding extra security layers”, but omits to recognize the filtration. Neon’s fall does not erase the background question: what price does our intimacy have when artificial intelligence demands more and more data? The model to pay for calls can reappear in other forms and other markets, because the need to train systems will continue to grow. What happened in the United States is an early warning that we are not talking about science fiction, but about real proposals that already touch the user’s door. The decision, ultimately, is personal. Images | Xataka with Gemini 2.5 | Screen capture | Neon In Xataka | A new generation of robots promises precision and efficiency. It also opens the door to cyberspage risks

Alibaba is becoming the Ai Open Source sponator. Your family of Qwen models is putting the market above

by

The Chinese giant Alibaba has launched Officially QWEN3-OMNI, an open source artificial intelligence model that can process text, images, audio and video simultaneously. In fact, it is the first model that unifies these four modalities natively and does it completely free, something that none of its US competitors offers. Bet on the Free Code. While Openai and Google charge for using their most advanced multimodal models, Alibaba gives theirs under Apache 2.0 license. This means that any company can download it, modify it and use it commercially without any cost. This open source approach It is the trend that multiple Asian giants are adopting to cause global interest in their language models and that multiple developers around the world want to contribute to their evolution. It is part of China’s strategy to remain relevant in the AI ​​career. Image: Alibaba What can you do exactly. As points The company, QWEN3-OMNI simultaneously processes text in 119 languages, recognizes voice in 19 languages ​​and can speak in 10 different languages. Its “thinker-speaker” architecture separates the reasoning of the audio generation, promising real-time responses with latencies of just 234 milliseconds for audio and 547 milliseconds for video. Benchmarks. In 36 reference tests, QWEN3-OMNI exceeds open source models in 32 of them and establishes new general records in 22. In advanced mathematics (Aime25) obtains 65 points compared to 26.7 of GPT-4O. In writing tasks (Writingbench) 82.6 points, exceeding 75.5 GPT-4O points. While it is true that it is not being compared to Openai’s most avant-garde model to date (GPT-5), it is a real achievement what giants like Alibaba are doing with their free and open source models. Strategy. Alibaba is running a risky but intelligent play: democratize the multimodal AI to gain market share. “This could bring some changes to the panorama of the OMNI open source models,” explained The Qwen team. The announcement occurs just when Nvidia announces Investments of 100,000 million dollars in data centers for OpenAI, while Alibaba and the rest of Asian giants prefer to dispute technological leadership in AI from another angle. What does it mean. Great American technology have opted for proprietary models that generate direct income. Alibaba wants to change the rules by giving instant access to its technology to millions of developers. Even if they offer it for free, they are building an ecosystem that gives them competitive advantage In the long term. And now what. China is not the only one that launches free code models. OpenAi has GPT-Oss And Google has Gemma. Two options that developers have on hand to deploy their ideas, modify them, contribute to their evolution and others, although they are not the main approach of both companies. In the case of Alibaba models, Deepseek either Tencentthe idea does revolve around the open source, and the pulse does not tremble when offering their most powerful models for free (despite the fact that some more complete and specific options are reserved for special agreements). QWEN models A great reputation have been carved Throughout these last years, and this new evolution in his family marks a new ribbon for the rest of the companies, not only in efficiency, but in the deployment of this business model. Cover image | Alibaba and Growika In Xataka | Eight people. An hour of work. A budget dollar. 5,000 new podcasts thanks to AI

This is how new models differ

by

Mark Zuckerberg says Within five years we will use smart glasses And less the smartphone. We do not know if it will be so, but what we do know is that Goal wants to be the leader in this wearable category And they just made it even clearer with new intelligent glasses models, three to be exact. This is the catalog of glasses with the finish line. Goal Ray-Ban Display They are the most advanced model of the catalog and solve one of the shortcomings that had been dragging smart glasses, add visual information. The new ones Goal Ray-Ban Display are The first ones that have an interior screen that only you can see. On the screen, which is in the right lens, information is projected as maps if you are following a route, a video call, preview the photos we take with the glasses and even for real time translation. There is not everything left, the goal Ray-Ban Display They also come with what Meta calls “neural bracelet”. They work through electromyography and are able to translate muscle activity into concrete actions. For example, if we make the gesture of clamp with the thumb and the index it will be like clicking with the mouse, while if we use the thumb and the heart the action will be to go back. At the moment the new goals with finish screen They will only be sold in the United States, where they will cost $ 799 and arrive in very limited units. It is expected to reach more countries in early 2026, although for now we do not know if Spain will be among the chosen countries. Ray-Ban Meta 2nd Generation Are the successors of the Ray-Ban Meta original As for the design, the first generation was available in the Wayfarer and Skyler models, but the new one also arrives in the Headliner model. The second generation improves some key features such as the camera, which now is capable of recording in 3k and its battery lasts double. According to goal, up to 8 hours with a single load. They are only two improvements, but their impact on the price is quite remarkable. While the first generation had a starting price of 329 euros, the new amounts to 419 euros. Oakley Meta Vanguard Are the successors of the Oakley Meta HSTNa model with sports aspirations, but a design too ‘Lifestyle’ that left them in no man’s land. The new ones Oakley Meta Vanguard They bet on a much more aggressive look and a Sports design that adapts to the face. You can choose different colors, the bad thing is that they do not offer the option to wear graduated lenses. It has a 12 megapixel panoramic chamber and records 3K video to 30 frames per second or FullHD at 60 frames per second. The speakers located in the pins have also been improved and target states that They have been optimized so that we can hear them well even if there is wind. The battery reaches 9 hours and also have a load case that can extend autonomy up to 36 hours. Beyond design, glasses also have functions oriented to sports such as Integration with Strava and Garmin. They can also be connected to Apple and Android health to obtain summons of training directly in the Meta AI app. If you also have a Garmin watch, you can ask the glasses details about the training, such as your heart rate. The price is much higher, specifically 549 euros compared to the 439 euros that cost the Oakley Meta HSTN. Meta smart glasses specifications Finally, we leave you a table with the key specifications of the models of glasses with AI that target has in its catalog, including the ones we already knew. For the first and second generation Ray-Ban we have taken as a reference for dimensions and weight the Wayfarer model. Ray-Ban Meta 1st Gen Ray-Ban Meta 2nd Gen Goal Ray-Ban Display Oakley Meta HSTN Oakley Meta Vanguard weight Glasses: 48g standard, 50g large Case: 133g Glasses: 51g Standard, 53g Grande Case: 133g Glasses: 69g standard, 70g large Bracelet: 42g Case: 169g Glasses: 53g Case: 213g Glasses: 67g Case: 258g Crystals supported For sun in several tones Transitions (they darken alone) Transparent Graduated lenses For sun in several tones Transitions (they darken alone) Transparent Graduated lenses For sun in several tones Transitions (they darken alone) Transparent Graduated lenses Prizm (mirror) Transitions (they darken alone) Transparent Graduated lenses Prizm (mirror) AUDIO Double speaker (76.1db) Five microphones Double speaker (76.1db) Five microphones Double speaker (76.1db) Six microphones Double speaker (76.1db) Five microphones Double speaker (82.1db) Five microphones camera 12MP ultra angular FullHD 30FPS video 12MP ultra angular 3K 30FPS Video, FullHD 60FPS 12MP ultra angular FullHD 30FPS video 12MP ultra angular 2203 × 2938 pixels at 30fps AUTONOMY 4 hours 36 hours with load case 8 hours 48 hours with load case 6 hours 24 hours with load case Neural Band up to 18 hours 8 hours 48 hours with load case 9 hours 36 hours with load case MEMORY 32 GB 32 GB 32 GB 32 GB 32GB others IPX4 resistance IPX4 resistance IPX4 resistance Monocular screen, 600×600 px, 90 Hz Neural bracelet with gestures using EGM IPX4 resistance IP67 resistance Compatible with Strava and Garmin price From 329 euros From 419 euros From 700 dollars, just use From 439 euros From 549 euros Images | Goal In Xataka | After mobile phones, cars, robots and AI, the next great technological avalanche of China arrives: the glasses

There are users selling their Wrapped data to train AI models. Spotify has not taken it well

by

Did you ever think you could make money with your Spotify data? What for many is only an annual summary with favorite songs and artists, for others it is a source of real value. An increasing group of users has decided sell your listening history To train artificial intelligence models. Spotify, which has turned Wrapped In one of its most recognizable brands, it has not taken long to react and the tension is served. Wrapped was born in 2015 as a tool to visualize listening habits and became a global tradition. Today, its popularity has given rise to UNWRappeda collective managed by the decentralized platform Vain. More than 18,000 users have grouped their data to sell them to interested companies. According to Ars Technicain June they achieved their first sale: a portion of data for $ 55,000 to only AI, with an average payment of about five dollars in cryptocurrencies for each participant. Wrapped is no longer just marketing: it is the center of a data control fight The interest in these data is not accidental: Spotify collects a detailed history of each account, searches, devices, approximate locations and technical records. All of that is available to any user through a Official download option, which offers packages in JSON format with years of history and precise data on the use of the platform. But although access is legal, its use has clear limits in the terms of Spotify. The policy for developers It prohibits using Spotify data or contents to train artificial intelligence models, resell information or replicate essential functions of the platform without permission. Spotify confirmed to the aforementioned medium that he sent a letter to those responsible for UNWRAPPed warning that the project could violate its registered trademark and violate these policies. The company insists that users can download their data, but does not authorize that they are monetic through third parties. UNWRAPPED pagethat until recently allowed to register and sell data, is no longer available. At the time of publishing this article, an English message appears that we have translated: “This service is no longer available. We regret the inconvenience. If you have any questions, get in touch with support. ”It is not clear if the closure responds to Spotify actions or a decision of the developers, who have not given explanations about this. Unwrapped was born, according to its creators, so that users had more control over their data, although their future is uncertain. Vana, the platform that drives it, argues that the project allows to group information in a community and collectively bargain its use. They explain that it is unlikely that a user manages to sell their data separately: companies are looking for broad sets, such as those who create vain, and then distribute the income among those who participate. Wrapped continues to be for most users a curious and shared annual summary. But the interest in monetizing this data makes it clear that the discussion about who controls personal information is not yet resolved. Spotify has expressed reservations on projects that use their services out of what is allowed, while decentralized proposals try to gain relevance. It is about to see what will happen finally with this project and others of similar characteristics. Images | Spotify | Xataka with Gemini 2.5 In Xataka | Google has resolved the dilemma of the age on the Internet as disturbing possible: spying on to protect you

You can make money with your GPU when you don’t use it. It is enough that you lend it to those who train AI models

by

To execute and offer tools based on generative artificial intelligence, a lot of calculation power is needed (and that leads to a lot of energy). Therefore, the most powerful market cards and specific processors for Datacenters They are so quoted today, hence companies such as Nvidia, which specializes in this market, are reaping such an overwhelming success. And since not everyone can afford a powerful graphics card to experiment with AI, there is a service that we see more and more common: to rent a graphics card to remove an extra money. There are several platforms to get it and under these lines we tell you everything you need to know. How the business works. The model consists of acting as a host in a Marketplace where clients are looking for GPU instances for their AI projects. You set the price per hour, the platform manages payments and the client executes their work in an isolated container on your machine. You could say that it is like an Airbnb, but focused on computer hardware. Instances with an rtx 4090 in vast ai Numbers that we must take into account. An RTX 4090 is usually Between about 0.20 and 0.60 dollars per hour in these marketplaces, depending on the demand. In the best theoretical scenario, operating 24 hours a day for a full month, a high -end GPU could invoice around 240 gross dollars monthly (considering that we put it for rent 24 hours a day). But reality is usually more modest, since we have to discount what we pay on our electrical bill, the platform commissions (which can reach 24% in Platforms like Runpod) and, above all, that real occupation is rarely 100%. Expanding market. The price difference between traditional cloud giants (AWS, Google Cloud) and these P2P marketplaces is considerable. While renting a GPU on AWS can cost three or six times more, platforms such as Runpod or Vast AI offer access to very powerful graphics cards, as is the case of RTX 4090, for a few cents the time. And of course, these prices are really attractive to developers who want to experiment with artificial intelligence but do not have means to have a team comparable to the projects they work on. What you should know before starting. Turning your PC into a rental server is not plug-and-play. In most cases you need Install Linuxconfigure updated NVIDIA drivers, open network ports And keep your team working for the hours for which you have committed to offer it, together with adequate refrigeration, which will be necessary if your GPU is going to start working much more and for much longer. In addition, your customers expect the machine to be available when they hire it, which means that you will not be able to use it for gaming or personal work. It should also be noted that the income generated is also subject to taxation and it is possible that it is required to register as an economic activity in cases where income exceeds a certain threshold. There are certain risks. Beyond the wear that the hardware can receive for being constantly working, there are maximum performance, there are some security concerns. Although platforms use containers to isolate workloads, some experts warn about possible Vulnerabilities in multi-tean environments (those environments that serve several users) that could compromise our data or use the GPU to improper purposes. Is it worth it? For most users with a single GPU, the benefits are modest once all expenses and others are discounted. Now, the business makes more sense if you already have the amortized hardware, do not pay too much on your electrical bill and accounts with certain technical knowledge to maintain the stable system. Even more if you have a potential graphics card or level for datacentes. As an experiment or complementary income experiment it can be interesting, but do not expect it to make you rich. First steps. If you want to try it, start with offers “interruptibles“, that is, cheaper but that can be canceled, in order to know the real demand. Vast.ai and Runpod They offer detailed documentation to become host, including step -by -step guides and preconfigured templates. Of course, it is advisable to always control real electrical consumption and establish operation limits to prevent your equipment from becoming a slave to the background processes. Cover image | She Don In Xataka | Nvidia, TSMC and SK Hynix are the most powerful chip companies on the planet. None can allow any of the others to fall

How tiny models are taking the colors to the mastodons of the AI

by

AI does not seem to advance much. At least the “big” AI. The best market models are barely managed to make relevant qualitative leaps, and that despite being gigantic and the money, time and talent that companies invest in creating them. We have seen it with Call 4Claude 4 or the recent one (and disappointing) GPT-5. But while Esots gigantic models are less and less surprise, diminutos models are getting more and more. Something (small) is moving. Google moves file Fichita. Last week Google surprised us all with the launch of a small AI model. Well, not: tiny. It could almost be said that it is a “nanomodel”. Gemma 3 270m It is an extremely compact version with only 270 million parameters. How small is that? It is easy to understand when we compare that model with one of the most reputable Open Source models: Call 4: In its Behemoth version, 288b (1,066 times larger) Qwen 3 235b (870 times bigger) Deepseek R1 671B (2,485 times larger) A hyperefficient model. Google’s own ones made it clear that this model cannot compete with the great AI models, but that was not their goal. Its objective is to be hyperefficient and, attention, hyperspepecific. What is pursued here is to turn Gemma 3 270m into the pillar of many models adapted to very specific and concrete tasks. The secret is called fine adjustment. GEMMA 3 270M, insisted these engineers, is a perfect model for fine adjustment processes (Fine Tuning) in very specific tasks. A company (or developer) anyone can take a small model, like this, and train it with their own data and refine it For a specific task following Google instructions For Hugging Face. For example, for generate stories to read children at night (code), to convert confusing text into structured data, to Customize messagesto classify emails or support tickets, or even for Play chess decently. Small models to power. Google already opted for this type of small models when GEMMA presented 3 In March. At that time the versions presented were 1b, 4b, 12b and 27b, being the last one really “great” in absolute terms. The rest could be executed at local in machines with 16GB of graphic memory, such as a Mac Mini M4. It is precisely what we could check with GPT-Oss-20B (the download is about 12 GB), the Open Source model recently launched by OpenAi that behaved remarkably. But even the latter could be considered “great”, and in recent weeks and months we have seen more and more “tiny” models that encourage the market. Gemma 3 270m’s performance is surprising despite its small size. And yet, the best of all is not that: it is his ability to adapt it to a specific task. Examples everywhere. Microsoft has already opted for this type of models with Phi-3 and Phi-4 (14b), which in its launch competed with the Chinese model QWEN-2.5-14B, although these models again tried to raise “mini” alternatives to large models such as GPT-4O or call-3.3 70b. They could be used for fine adjustment, but they were already trained to adjust to various scenarios. Others, more unknown, have gone further: the startup liquid launched A model aimed at visual environments called LFM2 with only 440m parameters, and Nvidia has just launched Nemotron -Nano -9bthat achieves improve QWEN3-8B performance in various benchmarks. Perfect for mobiles and smart watches. Another advantage of these models is that thanks to their small size are able to run on many more devices, however modest. They are ideal to be able to be used for example in our mobiles, smart watches or even more limited products. Its efficiency is the order of the day: how Google highlightedin a pixel 9 pro a quantized version (int4) of Gemma 3 270m can manage 25 “conversations” (chats) using only 0.75% of the mobile battery. It is so small that it can even be executed in a browser tab as if we load a website (heavy) more, such as the example of the web application that generates sleeping stories to children or This other which shows us how the model begins to break, but in a fun way, and whose code is available. A promising future. The Google model, like similar ones, raises that other side that Google spoke. More than an off -road model, what they offer is a base on which to build “the right tool for work.” These types of small models, well refined and trained, can be the basis of the design of all types of small applications and AI agents that then end up interconnecting and that function very, very efficiently. Perhaps it was true that the best essences are sold in small bottles. Image | Amanz In Xataka | If the question is which of the great technology is winning the AI career, the answer is: None

five big, light and cheap models to study

by

The new Aliexpress campaign is leaving discounts on all types of products: from video games and consoles Even tablets to face next school year. If you are looking for a good tablet to study, During the campaign of return to school there is a lot to choose fromso in this article we have reviewed some of the models with the best price. Xiaomi Redmi Pad 2 by 118.99 euros With the Esbs25 coupon, a tablet that offers 2.5K resolution and is compatible with Dolby Atmos. Little pad by 167.60 euros With the SBS35 coupon, a model with Snapdragon 7s Gen 2 processor and with microSD card slot. Xiaomi Redmi Pad Pro by 176.32 euros With the Esbs35 coupon, a tablet compatible with Dolby Vision that supports a fast charge of 33W. OnePlus Pad 2 by 299.75 euros With the Esbs55 coupon, a tablet with the Snapdragon 8 Gen 3 processor that supports a 67W fast charge. OnePlus Pad 3 by 453.71 eurosa 13.2 -inch screen tablet, Snapdragon 8 elite processor and fast charging of 80W. It is worth mentioning that the store has a large assortment of offers to which a discount coupon. If, on the other hand, you are looking for another device beyond the tablets, these are the coupons you can take advantage of: Discount Minimum purchase coupon 2 euros 10 euros Esbs02 BSES02 3 euros 15 euros Esbs03 BSES03 5 euros 30 euros Esbs05 BSES05 10 euros 69 euros Esbs10 BSES10 15 euros 99 euros Esbs15 BSES15 25 euros 139 euros Esbs25 BSES25 35 euros 199 euros Esbs35 BSES35 45 euros 269 euros Esbs45 BSES45 55 euros 349 euros Esbs55 BSES55 60 euros 429 euros Esbs60 BSES60 100 euros 599 euros Esbs100 – Xiaomi Redmi Pad 2 If you do not want to spend a lot of money and you are looking for a tablet that is economical, but that allows you to study and take it to class, the Xiaomi Redmi Pad 2 has dropped in price in Aliexpress to the 118.99 euros When using the coupon Esbs25 Before processing the purchase. The Xiaomi Redmi Pad 2 It is a tablet that has a standard 11 -inch size and weighs 510 grams. Your IPS panel offers a 2,5k resolution And a 90 Hz refreshment rate, its battery supports fast charging of 18W and internal level comes with the MediaTak Helio G-100 Ultra processor. It also stands out for its speakers, which includes four and are compatible with Dolby Atmos. Xiaomi Redmi Pad 2 – SBS25 coupon * Some price may have changed from the last review Little pad A step above we find another model that is ideal to study and take to class: it is about the Little pada tablet that has a larger size and some specific specifications than that of Xiaomi. Its price, in this case, stays in 167.60 euros When using the coupon Esbs35. The Little pad It is a tablet that mounts a good 12.1 -inch screen with 2,560 x 1.600py Snapdragon 7s Gen 2 together with 8 GB of RAM and 256 GB of internal storage (card expandable MicroSD). In addition, its battery supports a quick 33W charge and has a weight of 571 grams. * Some price may have changed from the last review Xiaomi Redmi Pad Pro Almost for the same price that little PAD has, we find an alternative that is also quite interesting: the Xiaomi Redmi Pad Pro. It is a very complete tablet that can now be bought in Aliexpress by 176.32 euros using the coupon Esbs35. The Xiaomi Redmi Pad Pro It is a tablet that comes with a 12.1 -inch IPS screen and weighs 571 grams. Its panel offers a soda rate of up to 120 Hz and is compatible with Dolby Visionits processor is the Snapdragon 7s Gen 2 and its battery supports fast charge of 33W. Xiaomi Redmi Pad Pro – Coupon Esbs35 * Some price may have changed from the last review OnePlus Pad 2 If you can allow yourself to spend a little more, in Aliexpress we find one of the options that have the best value for this list: the OnePlus Pad 2 It is an excellent tablet that right now has dropped up to the 299.75 euros using the coupon Esbs55. The OnePlus Pad 2 It is a tablet that stands out in several aspects: its IPS screen is 12.1 inches, it is compatible with Dolby Vision and offers up to 144 Hz. Its processor is the Snapdragon 8 Gen 3comes with 12 GB of RAM and 256 GB of internal storage. Its battery supports Fast charging of 67Wweighs 584 grams and comes with a six speakers system. OnePlus Pad 2 – SBS55 coupon * Some price may have changed from the last review OnePlus Pad 3 On the other hand, if what you are looking for is a more recent and much more complete tablet, the OnePlus Pad 3 It is also an excellent purchase option that can be found on offer in Aliexpress. Does not admit any coupon, so it only has a direct discount that leaves it for 453.71 euros. The OnePlus Pad 3 It is a tablet that can serve as an alternative to a laptop. Mount one 13.2 -inch screen which offers a resolution of 3,392 x 2,400 pixels and refresh rate of up to 144 Hz, its processor is the Snapdragon 8 Elite And its battery supports fast charging of 80w. In addition, it includes a system of eight speakers, weighs 675 grams and has 12 GB of RAM and 256 GB of internal storage. * Some price may have changed from the last review Some of the links of this article are affiliated and can report a benefit to Xataka. In case of non -availability, offers may vary. Images | Javier CriminalvaXiaomi, little, OnePlus In Xataka | Best tablets (2025). Which to buy and 8 recommended models for all pockets and needs In Xataka | BEST TABLETS IN QUALITY … Read more

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.