Top
Best
New

Posted by piotrgrabowski 1 day ago

M5 Ultra Mac Studio Review(www.macstories.net)
256 points | 252 commentspage 3
devy 1 day ago|
This dream machine costs over $15k (not including the Apple Studio Display)? Nah, that dream is SO OUT OF TOUCH!
aenis 21 hours ago|
The irony here is, thats hobby hardware. You spend 20k and can run slow hobby models that are barely capable of anything unsupervised.

Entry level serious hardware starts at 100k, and a bit better but still almost-useful grade is 200k (8x rtx pro, plus a nice epyc pairing). Thats the sort of thing a salaried expert lets their employer buy them for sort of serious work.

Anything really serious is well north of 1M - not including the housing and commercial grade mains connection. And at best that buys fast Kimi K3 or GLM.

apexalpha 8 hours ago||
You can run Deepseek 4 flash on a 128gb Mac, comfortably with lots of cache and context on a 256gb one.

These models are not 'barely capable', they're contemporary near frontier.

aenis 5 hours ago||
I do have a 128gb mac, and the experience is, for my use cases, hobby grade. I use models on a daily basis for design work (chiefly), and sometimes implementation work. I can't justify a drop in output quality and significantly increased waiting time. Sure, if frontier did not exist, I'd use them, but way differently than fable or astra.
apexalpha 4 hours ago||
If I had 0 privacy concerns I could just use deepseek directly in the cloud but the whole point is to not put your data on someone else's computer.
12kaj2 23 hours ago||
The Year Of Local AI will be here no later than 2040, coinciding with the Year Of The Linux Desktop.
prmoustache 23 hours ago|
The year of linux on the Desktop was 26 years ago for me.
saejox 22 hours ago||
i can buy a house with that amount of money. it used to be car money.
villgax 23 hours ago||
Lol, try generation of images & videos on these, they ought to improve perf on Deep learning not just llms
vivzkestrel 13 hours ago||
- it ll take you 64 yrs to recover that cost

- https://sunkcost.ai/s/mac-studio-m5-max-128/qwen3.8-27b-q4/?...

slashtom 22 hours ago||
Fantastic review, this is how it should be done with local AI.
sghiassy 1 day ago||
Imagine spending a trillion dollars on data centers and then reading this article. Nightmare fuel for OpenAI
whalesalad 1 day ago||
For 99.99% of people, spending 15 grand on a Mac Studio just to run Qwen 3.8 locally is a non starter.
jmull 1 day ago|||
It's not the M5 Ultra itself, but the M7s or M9s that will do the damage.

99% of people will use whatever AI is free. The sophisticated, heavy users that are willing and able to pay a lot of money the ones that will be interested in controlling their inference bills.

Today, the sweet spot where an M5 Ultra makes sense is tiny. But we might expect that to grow a lot.

BatFastard 22 hours ago||
Anthropic is reporting 100 Billion ARR.

Even if you could get a frontier model, you would not be able to run it on any Mac. So speculating on what M7 or M9 will achieve in 5 years (if we even still exist) seems pointless.

sghiassy 22 hours ago||
Do you need a frontier model to write emails, check your calendar, search the web?

I don’t think Apple is going to lie down and cede AI to the cloud.

geodel 21 hours ago|||
> Do you need a frontier model to write emails, check your calendar, search the web?

How about writing mail to President and senators on AI doomsday scenario if frontier labs do not pace themselves?

That mini model on mac mini would scared to hell to do such thing. It need that rugged frontier model to speak truth to power.

BatFastard 20 hours ago|||
True, but do you need a 10,000 dollar machine to do so?

I would love to find an excuse to buy a 10,000 dollar machine! But I cant find one yet. My current cloud bill is in excess of 400 USD per month. Just can't achieve frontier model capabilities locally.

sghiassy 13 hours ago||
I think in 5 years (aka before OpenAI can pay off all its debt) a lot of AI will be running on your iPhone
Danox 16 hours ago||||
Not for some of those well-connected high school, college kids, whose parents are rolling in it, the next generation of talented Steve Jobs, Bill Gates, Zuckerberg’s and others are coming up… Guess what they’re gonna get for Christmas?
sghiassy 23 hours ago||||
Yes, but in 7 years?
whalesalad 22 hours ago||
In 7 years we will probably all be living under ground fighting skynet with plasma rifles made out of old microwave parts
sghiassy 20 hours ago|||
I’ve watched Terminator 2 extensively. I am prepared
fragmede 22 hours ago|||
You will. Some of us are going to be already ground into dust that the microwave parts are made out of. Others will be locked into our communism cubes with our daily allotment of entertainment and sustinece. let out into the sunlight for only 30 minutes per day.
beastman82 23 hours ago|||
at 15 tok/s
CamperBob2 22 hours ago|||
And nightmare fuel is just what they'll be selling at the UN this week, for this very reason.

Sam's address will probably be more riveting, imaginative, and terrifying than the last couple of Terminator screenplays. Legislators will lobby him to write the laws for them, and the ghost of Harlan Ellison will threaten to sue him.

ajross 21 hours ago||
I don't see how that math works? This is a $15k rig under benchmark and per the results it competes very acceptably against... one consumer GPU.

I really don't see who buys this, except people who want the Studio for some other reason. But nothing in the story says you want to fill racks with these instead of Blackwell or TPU parts; it's not even close.

sghiassy 20 hours ago||
Your math is correct, but it’s math based on today’s economics.

Think of a company like Apple moving onto your turf. They’re not going to cede AI to the cloud. They want their part of the pie.

So in 7 years, how much AI will be handled locally on your iPhone. And will you have repaid all the debt on your balance sheet before Apple eats your lunch

SXX 19 hours ago|||
I think most important thing is that Nvidia doesnt want to give 100% of market to frontier AI labs either.

It's way too easy for 1T+ frontier labs to ditch Nvidia. So Nvidia will also put effort to make sure there are open weights models and local hardware available.

And Apple will benefit from this too.

ajross 19 hours ago|||
There is zero chance that an LLM approximating a modern frontier model is going to be running on a phone in the next decade. Even if you grant that you could stack enough DRAM dies on top of each other in the package, that would be a three order of magnitude improvement in power efficiency just for the compute.
cptskippy 23 hours ago||
I think we'll eventually get to the point where folks will have a local AI agent but I think people need to temper their expectations to a degree. You aren't going to have data center level tok/s from a box sitting under your desk and you don't need instantaneous responses for many workloads. Having a local agent that can execute tasks over a couple days with your supervision that might otherwise take you weeks is perfectly acceptable.

However I also think that Agentic AI is very much not an out-of-the-box solution, local or otherwise, and it takes a high level of technical knowledge to create an effective AI agent. And there's a problem now where most orchestration is fixed on what models are used for what tasks with no ability to weight constraints like cost, speed, and security.

Danox 15 hours ago|
The solutions will come in the past, big iron, thought they were immune and not too long afterwards, personal computers took over and big iron was history well, the same thing is going to happen all over again.

Those companies building those big data centers are going to find out that they overspent. Because we’re not going back to the mainframe era, no matter how much OpenAI, Anthropic, Microsoft, Meta or Google would like to.

lowbloodsugar 19 hours ago||
Everyone looking at the Qwen3 27B model and the 5090. It’s like saying a Porsche is better than a $5m Komatsu earth mover at moving a 20lb carry-on suitcase. Yes. Yes it is. Why do people spend $5m on a komatsu then, when this one metric shows the Porsche is better? Huh. Show us the “Moving 300 metric tonnes in one load” metric. How’s the Porsche now? Oh, the Porsche is in the Komatsu? Ok I’m getting a bit carried away with that analogy.
rwissinger 1 day ago|
[dead]
More comments...