Tuesday, August 25, 2026

Mac Studio 2026

Apple (Hacker News, MacRumors):

Mac Studio with M5 Max features an 18-core CPU, an up-to-40-core GPU with Neural Accelerators built into each core, and up to 128GB of unified memory, accelerating complex pro and AI workloads. With the powerful M5 Ultra, Mac Studio scales up to a 36-core CPU, up to an 80-core GPU, and a staggering 512GB of unified memory, enabling users to run enormous LLMs entirely on device. Wi-Fi 7 and Bluetooth 6 come to Mac Studio for the first time, while Thunderbolt 5 rounds out its extensive connectivity, so users can take advantage of blazing-fast external storage, PCIe expansion chassis, and powerful hub solutions for the most intense workloads. Thunderbolt 5 also enables multiple Mac Studio systems to be clustered, bringing up to 3x faster performance for distributed AI inference when compared to a single system.

[…]

Mac Studio with M5 Max starts at $2,499 (U.S.) and $2,299 (U.S.) for education.

[…]

Mac Studio with M5 Ultra starts at $5,499 (U.S.) and $5,099 (U.S.) for education.

It’s not shipping until September 22, with the 512 GB configuration in late October.

John Gruber:

Here’s my attempt to put all of the RAM/SSD configurations into condensed tables, so you can see which storage and memory options are available for each chip, and how much they cost.

[…]

It kind of stinks that there are no RAM options for the Studio between the 96 GB base and the $4,000 256 GB upgrade.

Jason Snell:

The Mac Studio is effectively the replacement for the Mac Pro, and it’s alone in offering the Ultra-class chip. Apple has positioned the Mac Studio as ideal for local AI workflows, and the new models will still be able to cluster via Thunderbolt 5 to create even larger collections of memory and performance. Apple representatives pointed out that a four-Mac Studio AI cluster, like the one I saw on display at WWDC earlier this summer, is so efficient that it can be powered from a single standard wall outlet. (They even showed us a picture of four Mac Studios plugged into a power strip that was plugged into the wall.) In an era of data center excesses, Apple is clearly leaning into the possibility of small, efficient Macs being used to do local AI work rather than relying on huge, power-hungry cloud models.

Federico Viticci:

If these numbers hold up and scale linearly, a local Mixture-of-Experts model such as Qwen 3.5-35B-A3B, which would run at ~17 tokens/sec on average on a base M4 Mac mini with 16 GB of RAM, could realistically generate output at over 60 tokens/second with the base model M6 Mac mini.

Based on what we’ve seen so far, the one downside of the M6 Mac mini is that it does not have Thunderbolt 5 ports; those are exclusive to the M5 Pro model, also announced today.

[…]

Speaking from personal experience, I know that my M3 Ultra Mac Studio using oMLX can run DeepSeek-V4-Flash locally with generation averaging 35 tokens/second. Assuming a linear 4x increase, that would put the same model at over 120 tokens/second on an M5 Ultra Mac Studio. To put things in perspective, that kind of performance would be faster than any AI chatbot website, it’d be faster than many providers who offer a “fast” mode for their models, and it’d only be second to either dedicated NVIDIA PC clusters at home or specialized inference providers such as Cerebras or Groq…which are running in full-blown data centers. Sure, you would need a computer that is likely going to cost more than $20,000 to make it happen, but it’d still be possible on a single machine that is small, quiet, and that – in theory – any consumer can buy off the shelf.

Previously:

Update (2026-08-26): John Gruber:

That upgrade is labeled “+ $300”, which makes it look as though the starting price for the 18/40-core model is $2,800. That’s the price I put in the original version of my chart.

But if you select that option, you’ll notice that the actual starting price jumps from $2,500 to $3,100 — a $600 difference, not $300. The reason is that the $2500 18/32-core version only comes with one option for RAM: 32 GB. The 18/40-core chip has three tiers for RAM: 48, 64, and 128 GB.

Update (2026-09-02): Joe Rossignol:

Apple’s press release for the new Mac Studio with M5 Max and M5 Ultra chips last week initially stated that the computer had “next-generation SSD architecture built on PCIe Gen 6.” However, as spotted by the French blog MacGeneration, Apple removed the PCIe 6.0 mention from the announcement shortly after it was published.

16 Comments RSS · Twitter · Mastodon


I no longer have any conception of how fast these computers are. Even the slowest new computer is more than fast enough for anything I'll ever want to do. We used to get new CPUs with benchmarks for photo editing, and then video editing, and now we're basically having to invent the biggest possible computational problems we can imagine in order for the performance differences to mean anything.

Accounting for inflation, US$2000 in the early 2000's (the price of an entry-level PowerMac at that time) would be over US$3600 today (or about the price of a top-of-the-line Mac Studio M5 Max).


@tim what you're effectively paying for as "performance" has grown detected from most people's needs, is things you can do at once (RAM) and quantity of supported displays, both of which Apple now welds to the rest of the configuration, so it's "balanced" in favour of people paying Apple huge sums of money for capabilities they don't need, to get capabilities they do.


detected = detached... *sigh*

@mjtsai any chance of a timed edit feature? OSNews has it if you want to see in action. Once you post, you have a 5 minute edit window.


@Someone I don’t think WordPress offers that as an option, but I’ll make a note to see if there’s a good plug-in.


@mjtsai That's why I mentioned OSNews, because they're running on WordPress. It's a possibly a specific plugin.


@mjtsai while you're at it maybe you can have a cookie to remember Name / E-mail / Web site info so we don't need to keep filling out ;)


@Marcos Will look, thanks.


I've been waiting for the new Mac Studio to replace my 2019 27" iMac (Intel). At first, I was waiting because I obviously tend to keep computers for a while and wanted to get the newest model. Now I'm leaning toward using the Apple Upgrade lease program with smaller payments for 24 months versus the 12-month Apple Card option. Then I would have the option for a new Mac in 2 years, or to pay the balloon payment and keep it.


"is things you can do at once (RAM) and quantity of supported displays"

Are you listing these as examples of what people need, rather than CPU? Computers today are so fast and capacious that RAM hasn't limited me since I first upgraded to 16 GB, at least 15 years ago. And I had dual 30" displays on my old Mac Pro, and tried plugging in more displays but couldn't figure out how to make practical use of any more pixels. Even the cheapest Mac Mini has supported that many displays (and pixels) for well over a decade now.

What Apple holds over customers is operating system versions. They're the sole supplier of macOS, and if you want to run current applications and security patches, you need a fairly recent macOS, and for that you need fairly recent Apple hardware.

If not for macOS/app version limitations, I'd still be daily driving my 2009 Mac Pro. There's nothing I want to do today that it wasn't great at, except running >2015-era macOS. 4C Xeon, 16 GB RAM, 4x SATA SSDs, dual GigE, and PCIe graphics is still a perfectly respectable workstation, as long as you're not doing anything crazy like local LLMs.


@tim What I meant is what you're paying for with a Studio, Vs. a Mac Mini, and for a M(more recent) Vs a M(older) processor. More Ram, and a greater number of displays supported are the fundamental differences in each beyond "does this a bit faster".

Otherwise, it's incrementalism, that most people won't notice, or care about... from the company that was telling everyone an iPad was all the "computer" they needed.


The cool thing about the Apple Silicon era is I only need to look for the form factor I want with the storage I want. Performance of the slowest new machine is more than good enough.


*sighs loudly at Viticci's AI boner*


I think the excitement about using these for LLMs is warranted. Open-weight models like GLM-5.3-flash and Qwen3.8-Flash-Next are getting to the point where they can do 80% of my dev work. Running extremely good models on (almost) humanly affordable hardware is becoming a reality.

Which is funny for me, because I'm transitioning my non-work desktop OS to Linux, so pretty soon, I might say something like, "macOS is great for backend work, but I do not find it very appealing on the desktop."


> What Apple holds over customers is operating system versions. They're the sole supplier of macOS, and if you want to run current applications and security patches, you need a fairly recent macOS, and for that you need fairly recent Apple hardware.

This started bothering me as well lately. Recent announcements about Intel Macs, Watches and iPads don't make me feel too optimistic about my aging first and second gen Apple Silicon Macs.


@Plume

I get that. I'm just EXTREMELY fatigued from seeing 'AI' and LLMs thrown in every situation, as if there's no other thing you can use a computer for today.


Jason Anthony Guy was quoted over in the new Mac mini post: "Why announce (and take pre-orders) today, but not ship for a month? Perhaps Apple is using pre-orders to gauge interest and adjust its product mix. With huge prices for memory and storage, Apple does not want to build high-cost configurations that end up sitting around."
What surprised me a bit was that Apple didn't deliver a Mac Studio that could use LPCAMM2 as a Level 3 RAM cache. They had 18 months… seems to me Apple engineering either really missed the boat or they got blinded by the Marketing/Exec teams' greed. Because the absolutely BEST WAY FORWARD would be to use current paradigm RAM on SoC for some base amount of memory—16, 48, 96, 256—and then have an LPCAMM2 slot to add additional, lower-performance RAM that would sit between main memory and the SSD virtual memory. This is already done in higher-end server/workstation configurations, albeit without the LPCAMM2 (they just use DIMMs). We're already seeing significant performance gains with some of the bigger LLMs just moving lesser-used parameters to SSD, having a SoC-based memory manager (that apps can influence) with Level 3 cache RAM would solve this problem. And at the level of the Studio, since the Mac Pro is gone, it really makes sense. I don't know that it really is necessary at the M7 Pro level, or Mac mini Pro, though having the option on a MacBook Pro|Ultra would make sense. And it would keep >$2500 Macs out of landfills—which SHOULD BE a goal for Apple (Right, Apple??)—while also greatly alleviating supply chain constraints on sales/delivery times.

My beloved 'first Mac', a IIci, had the ability to install a Level 2 cache card that sat between the CPU and main RAM; now the performance opposite is true: SoC RAM is faster, but an interim step before NAND is called for, better if it is user-expandable based on use case. Apple has engineered the solution in the past… will they be so forward thinking?? I'm HOPING to see this with the M7 Max/Ultra refresh… if I don't, my 'belief' in Apple silicon engineering will crater.

Leave a Comment