- that a bunch of users won't jump ship if Apple stagnates for 5 years
- that a product based on a model with Q12026 SoTA performance would be competitive with products using 2031's models.
- that just having access to good (by 2025/2026 standards) models is the big thing that Apple needs in order for Apple Intelligence to finally be useful.
This is the most unlikely proposition, given how Apple has managed to be a decade behind and still very profitable.
Besides that, having a Claw-like AI with full access to your phone is surely a recipe for disaster. IMO Apple is being justifiably cautious in staying an spectator, looking busy, and waiting to make a deal with the winner of the chatbot wars.
And concerningly, their AI products, like their search, will be powered by their biggest competitor in the mobile space. If AI becomes a killer feature, Google has the power to cripple the iBusiness simply by not renewing their current deal. At the least they should have adopted the Microsoft strategy and gotten into bed with all other major AI players.
I get the sense Apple leadership was never bought into the AI revolution (cf their anemic product releases, some of the papers they published) and we're seeing a shakeup in their ranks because they've realized they made the wrong bet.
A good analogy would be streaming. To get good quality, sure, you can store the video file but it is going to take up space. For videos, these are 2-4GB (lets say) and streaming will always be easier and better.
For models, we're looking at 100s of GB worth of model params. There's no way we can make it into, say, 1GB without loss in quality.
So nope, beyond minimal classification and such, on-device isnt happening.
--
EDIT:
> Nobody wants to be sending EVERY request to someone else's cloud server.
We do this already with streaming. You watch YouTube that is hosting videos on the "cloud". For latest MKBHD video, I dont care about having that locally (for the most part). I just wanna watch the video and be done with it.
Same with LLMs. If LLMs are here to stay, most people would wanna use the latest / greatest models.
---
EDIT-EDIT:
If you response is Apple will figure it out somehow. Nope, Apple is sitting out the AI race. So it has no technology. It has nothing. It has access to whatever open source is available or something they can license from rest. So nope, Apple isnt pushing the limits. They are watching the world move beyond them.
I'm pretty sure in five years, local LLM will be a thing.
Unless we invent a completely NEW way of doing videos, there's no way you can get that kind of efficiency. If tomorrow we're using quantum pixels (or something), sure 500MB is good enough but not from existing.
In other words, you cannot compress a 100GB gguf file into .. 5GB.
100GB to 5GB would be 20x. Video has seen an improvement of that magnitude in the days since MPEG-1.
It's interesting to consider that improvements in video codecs have come from both research and massively increased computing power, basically trading space for computation. LLMs are mostly constrained by memory bandwidth, so if there was some equivalent technique to trade space for computation in LLM inference, that would be a nice win.
That's not a matter of Moore's Law failing, but short-term capacity constraints being hit. It's actually what you want if Moore's Law is to keep going. It's a blessing in disguise for the industry as a whole.
Single-threaded compute, maybe - but that's increasingly a niche. Highly parallel workloads are still going strong in the latest device nodes, and power use for any given workload is decreasing significantly.
It could work for Deadmau5 because it is probably popular enough to be part of the model. How about "Hey, play that $regional_artist's cover of Deadmau5" and the model needs to know about "regional_artist", the concept of "cover", where those remixes might be (youtube? soundcloud? some other place).
All of a sudden, it all breaks down. So it'll work for "turn off porch lights", but not for "turn off the lights that's in the front of the house"
This is a paradox right? Handset makers want less handset storage so they can get users to buy more of their proprietary cloud storage while at the same time wanting them to use their AI more frequently on their handsets.
It will be interesting which direction they decide to go. Finding a phone in the last few years with more than 256gb storage is not only expensive AF, its become more of a rarity than commonplace. Backtracking on this model in order to simply get AI models on board would be a huge paradigm shift.
Useful LLM usage involves pushing a lot of private data into them. There's a pretty big difference sending up some metadata about your viewing of an MKBHD video, and asking an LLM to read a text message talking about your STD test results to decide whether it merits a priority notification. A lot of people will not be comfortable with sending the latter off to The Cloud.
And anyway, you already see models like Qwen 3.5 9B and 4B beating 30B and 80B parameter models, which can already run on phones today, especially with quantization.
Benchmarks: https://huggingface.co/Qwen/Qwen3.5-4B
Or pull out the phone and ask "Who's the person I met on X day ..".
Indeed.
But they said 5 years. That's certainly plausible for high-end mobile devices in Jan 2031.
I have high uncertainty on if distillation will get Opus 4.6-level performance into that RAM envelope, but something interesting on device even if not that specifically, is certainly within the realm of plausibility.
Not convinced Apple gets any bonus points in this scenario, though.
Then why does my M4 run models at TOK/s that similar priced GPUs cannot?
Gains in model performance isn't exactly cheap, and once one frontier model figures it out, the rest seem to copy it quick. Let them figure out what works and what doesnt, then put the "Apple" touch on it, all while putting your devices in everyone's hands. That's been their business model for years.
For Private Cloud Compute specifically, the system is described as underpowered and perhaps more trouble than it’s worth. Updating the software is apparently trickier and takes time, and more fundamentally the chips (believed to comprise right now of modified M2 Ultra processors) are not powerful enough to run the latest frontier models like Gemini, which the new Siri will be based on.The local AI community would strongly disagree with that assessment. They may not be able to run them with low latency for interactive use and this is most likely the real blocker for them, but they will have strong compute per watt compared to nVidia GPU's.
> like Gemini, which the new Siri will be based on.
The local AI community isn't evaluating the internal Gemini models. Apple's Private Compute hardware is specifically competing against Google's TPU hardware, which is a foregone conclusion if you've seen the inference economics. The money and electricity wasted on Mac inference at that scale isn't even attractive to Apple.
That's why I use an iPhone. I don't need and I don't want any "AI" in my phone. The claims that people want it comes from CEOs, marketers and influencers of GenAI companies, not from users.
https://www.reuters.com/business/google-apple-enter-into-mul...
1/ No, you don't get Opus 4.6 level on devices with 12Gb of RAM, 7B quantised models just don't get that good. Still quite good mind you, and I believe that the biggest advance to come from mobile AI would be apps providing tools and the device providing a discovery service (see Android's AppFunctions, if it was ever documented well): output quality doesn't matter on device, really efficient and good tool calling is a game changer.
2/ Opus 4.6 is now Opus 4.6+5years and has new capabilities that make people want to keep sending everything to someone else's cloud server instead of burning their battery life
So unless iPhone 20 Pro Max has 100GB of unifieid memory all of this is just pipe-dream. I mean, it wont even have 32GB of unified memory.
A claim like that is at best naive, and in any case ridiculous
And that is exactly why it won't happen (like that).
https://www.cnx-software.com/2026/02/22/taalas-hc1-hardwired...