In a bold attempt to rival Apple's Mac Studio, Beelink, a Chinese company, has announced its new mini workstation, dubbed the AI Mini. This compact yet powerful device is designed to meet demanding computing requirements.
The AI Mini is Beelink's second mini PC powered by AMD's Ryzen AI Max+ 395 processor, which combines the Zen 5 CPU architecture with a high-performance integrated Radeon 8060S graphics card.
This 16-core, 32-thread chip delivers up to 126 TOPS of AI compute capability. Its APU architecture merges CPU and GPU processing into a unified platform, improving efficiency for both general-purpose computing and AI-specific tasks.
Huge RAM for handling large datasetsAccording to Beelink, the AI Mini can function as a local AI server, making it ideal for developers running intensive models like DeepSeek R1, without the need for external GPUs or cloud-based infrastructure.
The device supports up to 128GB of RAM, boosting its appeal as a high end workstation for professionals handling large datasets, video rendering, or machine learning workloads.
Beelink has confirmed the inclusion of dual USB4 ports, each capable of speeds up to 40 Gbps, along with a USB Type-C port located on the front panel. Additionally, the AI Mini supports dual 10 Gbps Ethernet ports, delivering seamless connections to multiple devices.
Priced at $1,999, this business PC is not inexpensive, at nearly double the cost of Beelink's previous SER9 HX-370. However, similar devices based on AMD’s Strix Halo chip, such as the GMKTec EVO-X2, have also crossed the $1,500 mark.
Via Beelink
You might also likeAirbnb has been a disruptor since it hit the scene in 2007. Now, at its annual Summer Release event, it’s redesigning its app to make navigating easier and ultimately leveling up what you might expect from a stay at one of its many properties worldwide.
For instance, while you get a bit more space with an Airbnb, you might lose out on some amenities, be it a gym, an attached restaurant, or even some self-care. The tech company's answer is dubbed Airbnb Services, essentially a way to book a service like a massage, a haircut, or even a chef to arrive at wherever you’re staying.
That aims to fill in the missing gap of what you might get from a hotel, but it also gives you, the visitor, a bit more control, with the option of looking through the available options and picking the ones that fit your needs the most.
Similarly, Airbnb Experiences aims to replace the mundane with memorable, allowing you to book activities from locals in the cities where you stay to get the most out of them. These seem especially fun, and at a launch event, Airbnb certainly brought out the talent to up this.
Then, tying this all together is a redesign app that looks a lot easier to navigate and acts more like a helpful companion. If you have a stay booked, as well as services or experiences, it can bundle them as an itinerary for you, listing important reminders.
1. Book a haircut, massage, or a catered meal with Services (Image credit: Airbnb)Services seem like the most approachable addition, and probably one of the most requested. Rather than finding out if the host approves of a third-party, Airbnb handles the vetting and will offer 10 categories of services at launch. Those include chefs, photographers, massages, spa treatments, personal training, hair treatments, makeup, nails, prepared meals, and catering.
It’ll launch in 260 cities across the globe at first and lives directly in the Airbnb app – there is an entire ‘services’ category at the top, and after you book a location, or are considering, you can see recommended services.
Airbnb is vetting the service providers and promises that each has around 10 years of experience in the respective industry. Additionally, each instructor or provider has to go through a verification process and list out applicable trainings or certifications.
(Image credit: Airbnb)Services will start at $50 in the US (we’re waiting for exact starting prices to be confirmed in the UK and Australia, but that converts to around £40 / AU$80) and are rolling out now. Airbnb is also encouraging providers to apply to become hosts for services on the platform.
The real kicker here, though, and one that potentially could have TaskRabbit concerned, is that you don’t need to book an Airbnb home for a stay to use Services or Experiences (more on this in a bit).
If you need a photographer for a birthday party or a shower, you could fire up Airbnb on your iPhone or Android phone, look through reviews, and hire a professional. Same for a hibachi chef for a girls' night out – at an Airbnb or in your own home – or for a caterer specializing in soup dumplings.
2. Memorable moments dubbed Experiences (Image credit: Airbnb)Next to Services up-top in the Airbnb app will be ‘Experiences’ with a hot-air balloon icon. And similar to the 'Icons' homes that Airbnb lets folks enter to win a stay in – like the house from Up or the control room from Inside Out – Experiences is a bit more pie-in-the-sky, in the best way possible.
These are unique activities run by folks from where you’re staying that aim to let you really immerse yourself in the locale. It could be a cooking class or a dining experience, but rather than just a meal at a restaurant, it’s interactive and potentially led by a Michelin-star chef, or you can freshen your wardrobe with a celebrity stylist.
(Image credit: Airbnb)Similarly, rather than just visiting a historical place, you can be led through the site with a dedicated historian or take a workout class with an expert in that specific field. The idea is to have true experts for a more compelling, well, experience.
Experiences will be launched in 650 cities worldwide, and Airbnb says that will be expanding. There will also be a higher tier dubbed Airbnb Originals, which the company promises could lead you to eating barbecue with Patrick Mahomes in Kansas City or playing volleyball with an Olympian. There’s no word on pricing for these, but I suppose these might be more similar to Airbnb Icons, in that you enter for the chance to use these.
More likely, you’ll find yourself with the more standard Airbnb experiences, which aim to be more than just standard.
3. App redesign (Image credit: Airbnb)Helping to make all of these more appealing, and honestly more integrated into the booking of an Airbnb – maybe a beach bungalow, a fancy loft, or a classic cabin – the app is also getting smarter in some key ways. The layout aims to make this flow a bit easier, and for the app to act as a companion serving up reminders and suggestions.
The biggest change, though, is that Airbnb now allows users to search for properties worldwide alongside the best services and experiences all in one place. Under the hood, Airbnb says it’s built an entire new tech stack that adds capacity for these new offerings and sets up a large runway for the future.
And I think that hints at the broader play here – Airbnb doesn’t want to be an app you open once or twice a year for a big vacation or trip. Rather, it’s there and is an app with more tools in its chest, allowing you to book more casual everyday services but also more memorable ones, whether you’re on the go or for a staycation.
Carolina Milanesi, President and Principal Analyst at Creative Strategies, told TechRadar, “Expanding into services makes a great deal of sense as it allows for a richer experience than the stay itself and can certainly add to it by adding a more visceral component to the stay – say you are using a personal trainer while you travel or organizing a photographer for your girls reunion weekend. The option of booking even when you are not staying shows they are trying to expand and deepen the relationship with a customer by having multiple touch points through the year.”
It certainly adds up, and this could grow Airbnb’s user base as well. We’re going hands-on with the new app, and maybe even an experience or two soon, so we’ll report back with more. But if you’re keen to give it a go, update to the latest version of the Airbnb app now.
You might also likeWindows 11 has a new preview build out and it comes with a big change: the introduction of AI to the Settings app in order to help Copilot+ PC owners find and choose the options they need.
This is a new piece of work Microsoft announced last week, but the company just said that it was coming to testing soon, at some stage this month. Well, that stage has now been set, and this ability has debuted in preview build 26120.3964 in the Beta channel.
That preview release comes with the full AI agent present in Settings, although of course, this is just the initial stages of testing. Also note that this is for Copilot+ PCs only, it won’t be applicable to normal Windows 11 PCs, as it leverages the peppy NPU that’s on board Copilot+ devices. (Microsoft also revealed fresh goodies inbound for non-Copilot+ machines, it should be noted too).
As I’ve covered before, the AI agent works via the usual query model. You type in what you’re looking for in Settings via a bar at the top of the panel using normal conversational language. For example, you could ask something like, “How do I put the YouTube video playing on my laptop onto my TV screen?” and the AI will find that relevant setting for screen mirroring.
The idea is to make tweaking settings or finding various capabilities a good deal easier, and with some options, AI will even recommended specific choices for you.
Another change to the Settings app in this preview build is a new hardware-related FAQ that provides info on your PC specs and how good they are. This nestles in the System > About panel within Settings, underneath the list of device specifications.
The questions-and-answers specifically relate to your PC, so if you only have 8GB of RAM for example, and you’re worried that might slow down apps – or moreover PC games – there’ll be a section on exactly what that memory loadout means for you in terms of expected performance levels.
Analysis: PC spec cards and AI accuracy (Image credit: Future)This is the first time we’ve seen this hardware FAQ appear in testing, though oddly, it’s not partnered with the related ‘PC spec cards’ (small info panels) that carry the individual details of your CPU, GPU, system RAM and storage (which have been spotted in testing before, minus the FAQ).
Presumably, these pieces of the puzzle are going to come together soon, delivering a useful extra for relative PC newbies who may not be so sure about what all these specs mean.
As for the AI agent, it looks nifty and offers plenty of promise, but there are a few notable catches here. Firstly, AI could get things wrong when it comes to the specific recommendations given for certain settings (but hopefully not in terms of the basic finding and displaying of relevant options).
So, we need to be cautious in that respect, and it’ll be interesting to see how accurate Windows 11 testers find the system, and what the initial feedback is – because this could be one of the best uses of AI Microsoft has found yet. Or the worst if it comes horribly off the rails, but Microsoft doesn’t appear overly ambitious here – yet – so that shouldn’t happen.
Secondly, only the English language is supported by the Settings agent for now. Finally, just Copilot+ PCs with Snapdragon processors are supported in testing to begin with, but Microsoft has said the agent will come to Copilot+ devices with AMD and Intel CPUs soon enough. Perhaps that’ll be the next preview release, then.
You might also like...Alibaba's Tongyi Lab has found a way to train AI search models without using real search engines, which it says can reduce search training costs by up to 88% compared to commercial APIs like Google.
In a paper entitled "Incentivize the Search Capability of LLMs without Searching," Alibaba explains how the development uses simulated AI-generated documents to mimic real search engine outputs.
Interestingly, Alibaba's researchers also note that using simulated documents can actually improve the quality of training, because "the quality of documents returned by search engines is often unpredictable" and risks introducing noise into the training process.
Alibaba will train AI search models on AI-generated documents"The primary difference between a real search engine and a simulation LLM lies in the textual style of the returned content," the researchers wrote. ZeroSearch can also gradually degrade the quality of documents in order to simulate increasingly challenging retrieval scenarios.
Of course, the key benefit to this technology is the significant cost saving available. Training with ZeroSearch's 14B model costs around $70.80 per 64,000 queries, compared with around $586.70 via Google's APIs. Costs are even lower for the 7B and 3B models, at $35.40 and $17.70 per 64,000 queries, and yet all three of the ZeroSearch models and the Google API method take the same amount of time.
However, Alibaba acknowledged that one, two, or four A100 GPUs are required for its ZeroSearch method, compared with no GPU requirement via the Google API method, which could present a negative impact in terms of sustainability, like energy consumption and emissions.
"Our approach has certain limitations. Deploying the simulated search LLM requires access to GPU servers. While more cost-effective than commercial API usage, this introduces additional infrastructure costs," the researchers concluded.
Still, challenging the reliance on expensive and gated platforms like Google Search APIs and reducing the costs could help democratize AI development even further.
You might also likeAfter a series of robotics announcements over the past few years, including its latest Vulcan robot which has a sense of touch, Amazon is testing next-generation models named 'Stow' and 'Pick', and their names say it all – they're designed to assist in the stowing and picking processes at fulfilment centers.
However, the company has acknowledged that current robot technology is in no place to replace human workers despite achieving promising results.
According to its performance testing, humans averaged 243 units per hour compared with robots, at 224 units per hour. The difference is small, but Amazon also highlighted some nuances.
Amazon's robots are good, but not human goodWhere the robots stand out is in consistency. "It was also found that humans had greater variation in stow rates: people can quickly stow many small items efficiently, but are slower with large items, crouching for lower bins, or when using a step ladder to reach the top bins," Amazon explained.
The company also observed how humans can perform multiple tasks at once with two hands, such as pulling out a storage bin with one hand and stowing an item with the other – something that the current generation of robots fails to do.
Still, the gap is closing, with Amazon's robots stowing at greater than 85% success at a similar stow rate to humans across more than half a million tests.
That said, there are some challenges with implementing robotics across Amazon warehouses, including damages caused by dropping products, inserting products into stowage bins, and crushing lightweight boxes.
As such, while full human replacement is not feasible yet, hybrid systems are proving to be highly effective, with more work needed on improving handling and reducing damages caused by robots.
"Deeper research into predicting complex item interaction may be required in scaling manipulation more general," Amazon's researchers added.
You might also like