Pocster

Right , now, let’s do this slowly I didn’t check every website on Earth . You need enough ram to be useful . 24gb once booted won’t leave a lot for a useful llm . local llm can even be installed on a pi for example . The keyword is useful - in essence you’d want 64gb min on a Mac as it uses unified ram . People run these things on much less . These units and above have all but disappeared. Ultra m3 ( go check ) on eBay ( assuming they have real stock ) are on for 5k base model prices . So that’s 1k more than Apple retail them for . Openclaw etc has made a lot of people punt on a relatively cheap Mac mini to experiment with .

I’ll give you a 69

A sane person . Problem is unified ram . Best on Nvidia is 96gb on rtx6000 that’s a 10k gpu card !!!! This is why Apple have ( unintentionally) jumped to number 1 for local llm .

For completeness yes it’s on Amazon apparently at an inflated price .

Yeah it said “ out of stock “ so that’s a clue . Screen grab is from now !

Notice Mac minis basically out of stock and even a m3 ultra 96gb ( base model ) no where to be seen . I got mine from curry’s of all places ! , check it every few days to judge demand - now they have none ! I think when I sell the m3 I’ll get a pretty good price for it even with m5 released because it’s all going to sell out fast !

After lots of reading ( for probably months ) , YouTube videos etc etc . Enough to confuse anyone I decided to start Avalon . chatgpt and I are architects we setup llama etc , vs studio , ssh into m3 etc etc , even parsec screen share Deepseek is our reasoner . Qwen coder next our sweat shop “ just do it ! “ coder ! chat managed to create me some scripts so we create a brief for today’s task . Get Deepseek to take that and produce a detailed instruction list for qwen . So process is Create task Give to Deepseek Approve Deepseek output Tell qwen to do it ( for reasons I’m unsure of we don’t use git diffs but json - I didn’t argue as it seems to work ) Then some kind of test . Chat gets excited about this ; me less so when terminal window says “ it worked ! “ 😊 But we are building framework and structure first . I’m blown away by this all locally ! llm , creating document for llm to produce detail for another llm Crazy !! 90% of programmers obsolete ! Just todays tasks would have been weeks of learning and work .

I’ll use chat as my reasoner but I need local to do my local actioned stuffed until I can source enough ram . Bigger model did mean better but now smaller better optimised models can out perform a larger one . Obviously there’s a point when you can’t go smaller without too much compromise . If you ask ChatGPT itself how long before a model matches its capabilities of today in a ‘ reasonable ‘ size it reckons within 2 years a 256gb model will be comparable to ChatGPT 5.4. That’s insane ! To have that capability sat on your desk in a relatively low ram config. Moe and potentially’ bolt on ‘ specialists seems to be the next level . A medical specialist , maths specialist etc plugged in to the reasoner.

Even if Mac unified memory and iPhone RAM aren’t identical final packages, they come from the same constrained LPDDR supply chain. Apple then has to allocate that supply across products, and iPhone will obviously take priority over niche high-memory Mac configs.

TSMC doesn’t make RAM, no one said they do. TSMC is the SoC bottleneck. The RAM shortage is a separate bottleneck, and Apple clearly isn’t immune. Other manufacturers have openly said memory shortages are affecting them, and Cook has said Mac mini / Mac Studio supply-demand balance will take several months. High-memory Mac mini / Studio configs are also exactly where the worst delays and unavailable options have shown up. There are also multiple reports suggesting the memory shortage could last well into 2027 and possibly beyond, especially as AI demand keeps soaking up supply. So pointing out that TSMC doesn’t make RAM doesn’t disprove the RAM issue; it just identifies a different bottleneck.

🙄🙄🙄🙄😉

Well ! Unfortunately as I presume you now realise the ram shortage is a massive problem - yes , even for Apple . It did look obvious tbh - so now I have to wait for any indicator of ultra appearing and jump on it ! I guess it could be announced at wwdc - maybe have a date for pre orders . Buys Apple time to acquire ram whilst still be bigging up their product . That’s the best I can hope for - a definite date to refresh Apple Store !

I take your points, but I think the trend cuts both ways. Open source never really dies. So paying a modest annual fee for a strong local model — not the latest frontier model, but still very capable — seems perfectly plausible. I also don’t buy the idea that free models are just tasters. We’re in an AI arms race. Open releases help stop one company becoming too dominant, while building ecosystems, adoption and hardware demand. The very best models may remain closed, but strong local models being open or low-cost feels more likely than local AI fading away.

I suspect quite soon you’ll be able to buy the latest version for local llm for like £20 or something .

I’ll help ! Billions spent on llm / ai etc For a sum under 10k you can have something ‘near’ to that locally ! Thats insane !

Kimi K2.6 Wow ! I want that ! Write a custom iPhone app so I can chat directly to Avalon via tailscale !! Should effectively mean I don’t need a chat gpt + subscription as K2.6 is pretty darn near to chat 5.4 . My own local chat , accessible anywhere in the world would complete access to my entire home . What could possibly go wrong ?

9yrs ?

Yes Gemma is surprisingly good for a small model . But there is a limit to how small a model can go and still be useful / accurate . But even now a good optimised small model can outperform a larger unoptimised one . Biggest isn’t always best - well in llm model world anyway 😉

A reliable source - indicating the ram issues ARE a major factor . Gurman is seen as the number 1 journalist for Apple reporting . So this is a definite indicator of the issues affecting Apple . Man ! October !!!!

Fair points, and I agree Apple is probably too important a customer to be left without supply altogether. My concern is more that AI demand has already distorted the memory market. So even if Apple can get what it wants, it may be doing so at a much higher cost, and that cost is most likely to show up in high-end Mac pricing rather than in total unavailability. In other words, Apple probably still gets the RAM — the question is how brutally it prices the top bins, and whether the niche 256GB/512GB configs end up with longer lead times because Apple allocates supply where it makes the most commercial sense.

But AI cloud buyers have already distorted the RAM market. Apple may be a major long-term customer, but that does not insulate it from higher input costs. If memory prices are pushed up by AI demand, Apple will likely pass much of that on to buyers — especially on high-end configurations. And unlike most consumers, AI cloud operators can justify paying extreme premiums to secure supply. That can also reduce the amount of top-end memory available to Apple for niche high-RAM Macs, leading to longer delivery times or higher bins becoming temporarily unavailable.

Absolutely! . Apple could deliberately up their Apple ram tax even more . This would also push people to take a tier lower I.e I’d love 512gb but if price and delivery is horrible would take 256gb. Think I saw meta have also price increased due to ram shortages/costs on their vr system . My main concern is that AI cloud operators will pay whatever it takes to lock in RAM, while a relatively niche market like Apple’s high-end machines could end up squeezed as a result.

Yes we know that . But you have to factor in global ram supply issues . Come wwdc ( assuming released then ) we will see the real world impact of that on Apple . I think you mis understand me . I want you to be right ! I.e Apple have significantly buffered themselves against ram shortages ( bumped the price a bit already ) - that would mean supply of high memory configs on release day no issue . I want to be wrong . But it simply doesn’t look that way too me .

Lives too short to finish the build ! . I’ll try and do a bit when I get back

You get a (expletive deleted)ing photo for that !