Run DeepSeek R1 On An Apple M3 Ultra Mac Studio? Sure, It'll Just Cost You $9,499-plus

Apple's newly refreshed Mac Studio has arrived bristling with up to 32 CPU and 80 GPU cores, and as much as 512GB of unified memory on board.
Interestingly, Apple's top-specced Macintosh to date isn't packing the iGiant's home-grown Arm-compatible M4 processor — you only get that in the base model Mac Studio. Instead, the latest top-end Mac Studio features the long-awaited M3 Ultra.
Similar to the M1 and M2 Ultra, the new M3 Ultra is essentially two M3 Max processors that have been stitched together using TSMC's advanced packaging tech — or as Apple prefers to call it, "UltraFusion" — into a single logical chip with 184 billion transistors in total.
Compared to its predecessor, the M3 Ultra offers a substantial upgrade. The processor can now be configured with up to 32 cores (24 performance and eight efficiency) versus the M2 Ultra's 24 (16 performance, eight efficiency).
According to Apple this translates to a 50 percent performance uplift over the M2 Ultra and an 80 percent increase over the original M1 Ultra powered Mac Studio.
The GPU has also been upgraded from a max of 76 cores to 80 this generation. That might not sound like much of an upgrade compared to the M3 Ultra's CPU count jump, but, as you might recall from Apple's "Scary Fast" event all the way back in October 2023, most of the graphics improvements are actually under the hood, within the graphics cores rather than from upping the number of cores.
Apple has previously claimed the M3 GPU cores are about 80 percent faster than the M2's, thanks in part to the hardware-accelerated mesh shading and ray tracing. Add in the extra cores on the M3 Ultra and Apple says the GPU is roughly twice as fast in graphics workloads compared to the previous generation.
The GPU is complemented by a 32-core Neural Engine to accelerate all those Apple Intelligence features that are now turned on by default in macOS.
Speaking of AI, Apple is keen to tout the M3 Ultra's memory capacity which can be specced up to 512 GB, providing adequate space for "LLMs with over 600 billion parameters."
This might seem like a weird flex, but Apple's Mac Studio has become quite popular among AI researchers and enthusiasts as it's one of the cheapest platforms for running high-parameter-count large language models, aka LLMs.
And with 800 GB/s of memory bandwidth running something like DeepSeek R1 at home is totally feasible, at 4-bit precision at least. With 671 billion parameters, it'll eat up a little over 400GB of your memory, but because it uses a mixture of experts architecture, only 37 billion parameters are active at any given point. This means generation rates, potentially as high as 20 to 30 tokens a second, should be achievable by our estimation.
With that said, a Mac Studio capable of running such a model won't be cheap. Upgrading to 512GB will cost you an extra $5,500 over the standard M3 Ultra-toting Mac Studio, bringing the total to $9,499 before you even upgrade the storage. Fully maxed out, an M3 Ultra Mac Studio with 16TB of SSD storage, will run you a cool $14,099.
While Apple's top-specced Mac Studio gets the M3 Ultra, the base model comes equipped with the same M4 Max that's powered its MacBook Pros since last fall. That chip can be equipped with up to 16 CPU cores (12 performance, four efficiency), a 40-core GPU, and up to 128GB of unified memory. Additionally, storage is capped at a max of 8TB, though given the rather sizable premium Apple charges for storage, and the fact this thing will be sitting on a desk most of the time, we suggest investing in external USB4 or Thunderbolt storage, instead.
The M4 arrives on the MacBook Air
Alongside the Mac Studio, the venerable MacBook Air is also getting an Apple Silicon refresh. The passively cooled notebook is now available with an M4 processor and twice the starting memory while coming in $100 cheaper than the previous gen at $999.
As we discussed last year, the M4 offers a couple of notable advantages over its predecessors, but in the case of the MacBook Air, one of the most obvious is support for multiple monitors.
Unlike the M4 MacBook Pro launched last fall, the base model 13-inch Air comes equipped with a slightly cut-down version of the chip. Instead of the 10 CPU and 10 GPU core version that comes standard on the Pro, the base model Air's GPU features two fewer cores.
While the base SoC may have been cut down in terms of compute, its memory has been doubled to 16GB over the M3 Air's 8GB, though the minuscule 256GB SSD remains. Stepping up to a 10-core GPU and 512GB of storage will push the price to $1,199.
The 15-inch model, meanwhile, starts at $1,199 and comes standard with the 10-core GPU, 10-core GPU-equipped M4, and 16GB of RAM, but just 256GB of storage.
- iOS users left refreshing in vain as Microsoft Outlook woes drag on
- Apple drags UK government to court over 'backdoor' order
- Hey, remember iPads? Those fondleslabs? Apple still does
- Apple dares users to fix 'budget' iPhone 16e themselves
While Apple may have slashed the starting prices for the M4 Air, just like any other Mac, once you start upgrading the memory or storage — remember they're soldered down — they can get pretty expensive in a hurry. (There's a reason why we used to call it the Cupertino idiot-tax operation.)
A maxed-out 15-inch Air with 32GB of RAM, and 2TB of storage will set you back $2,400, at which point you're well into MacBook Pro territory.
Both the refreshed MacBook Airs and Mac Studios are available for preorder now and will hit store shelves on March 12. ®
From Chip War To Cloud War: The Next Frontier In Global Tech Competition
The global chip war, characterized by intense competition among nations and corporations for supremacy in semiconductor ... Read more
The High Stakes Of Tech Regulation: Security Risks And Market Dynamics
The influence of tech giants in the global economy continues to grow, raising crucial questions about how to balance sec... Read more
The Tyranny Of Instagram Interiors: Why It's Time To Break Free From Algorithm-Driven Aesthetics
Instagram has become a dominant force in shaping interior design trends, offering a seemingly endless stream of inspirat... Read more
The Data Crunch In AI: Strategies For Sustainability
Exploring solutions to the imminent exhaustion of internet data for AI training.As the artificial intelligence (AI) indu... Read more
Google Abandons Four-Year Effort To Remove Cookies From Chrome Browser
After four years of dedicated effort, Google has decided to abandon its plan to remove third-party cookies from its Chro... Read more
LinkedIn Embraces AI And Gamification To Drive User Engagement And Revenue
In an effort to tackle slowing revenue growth and enhance user engagement, LinkedIn is turning to artificial intelligenc... Read more