Apple Intelligence Foundation Language Models Tech Report 2025
machinelearning.apple.com
machinelearning.apple.com
> We do not use our users’ private personal data or user interactions when training our foundation models. Additionally, we take steps to apply filters to remove certain categories of personally identifiable information and to exclude profanity and unsafe material.
> Further, we continue to follow best practices for ethical web crawling, including following widely-adopted robots.txt protocols to allow web publishers to opt out of their content being used to train Apple’s generative foundation models. Web publishers have fine-grained controls over which pages Applebot can see and how they are used while still appearing in search results within Siri and Spotlight.
Respect.
https://www.bloomberg.com/news/articles/2024-06-12/apple-to-...
Also the bigger problem is, you can't train a good model with smaller data. The model would be subpar.
- Famous Dead Person
Robots.txt is already the understood mechanism for getting robots to avoid scraping a website.
Apple are saying you can opt out of their training data collection using robots.txt.
But... they collected their training data before they told people how to opt out.
I don't understand why me pointing that out as "eyebrow raising" is controversial here.
It might be nice if there were categories that well-behaved bots could follow, as noted above, but even then the problem exists for bots doing new things that don't fall into existing categories.
I think that's disingenuous of them.
Wanting to disallow LLM training (or optionally only that of closed-weight models), but encouraging search indexing or even LLM retrieval in response to user queries, seems popular enough.
An issue for anti-AI people, as seen on Bluesky, is that they're often "insisting you write alt text for all images" people as well. But this is probably the main use for alt text at this point, so they're essentially doing annotation work for free.
Computer vision is getting good enough to generate it; it has to be, because real-world objects don't have alt text.
Annotating photos takes time/effort, and I could totally imagine photo apps being resistant to prompting their users for that, some of which would undoubtedly find it annoying, and many more confusing.
Yet I don't think that one can conclude from that that annotations aren't helpful/important to vision impaired users (at least until very recently, i.e. before the widespread availability of high quality automatic image annotations).
In other words, the primary user base of photo editors isn't the set of people that would most benefit from it, which is probably why we started seeing "alt text nudging" first appear on social media, which has both producer and consumer in mind (at least more than photo editors).
One would hope they're responsive to user demands. I should say Lightroom does have an alt text field, but like phone camera apps don't.
Apple is genuinely obsessed with accessibility (but bad at social media) and I think has never once advocated for people to describe their photos to each other.
How did you come to the conclusion that those two groups overlap so significantly?
> this is probably the main use for alt text at this point
> Apple has since confirmed in a statement provided to Ars that the US federal government "prohibited" the company "from sharing any information,"
Take all the space you need to lay out your contrary case. Did the San Bernadino shooter predict this?
It is exactly the same thing as saying “if you ignore the heads, these coins really always come up tails”.
Does the Chewbacca argument method ever work these days?
They are decades behind in AI. I have been following AI research for a long time. You can find best papers published by Microsoft, Google, Facebook in past 15 years but not Apple. I don't know why but they didn't care about AI at all.
I would say this is PR to justify their AI state.
And then they just... gave up?
I don't know what happened to them. When AI breakthrough happened, I expected them to put up a fight. They never did.
Apple always had the luxury of time. They work heavily on integrating deeply into their ecosystems without worrying about the pace of the latest development. eg. Widgets were a 2023 feature for iOS. They do it late, but do it well.
The development in the LLM space was and is too fast for Apple to compete in. They usually pave their own path and stay in their lane as a leader. The impact on Apple's brand image will be tarnished if Google, Meta, OpenAI, MS all leapfrog Apple's models every 2-3 months. That's just not what the Apple brand is associated with.
Tim Cook happened. The fish rots from the head down.
The predecessor is no reason for the current head to slack off.
So, it's nice to see Apple is doing research and talking about it, but we're out here waiting, still waiting, for anything useful to make of it all on our thousand-dollar devices that literally connect us to the world and contain our entire life data. It's what I would've expected from one of the most valuable companies in the world.
Probably wouldn't have made a difference but the second half of that statement isn't exactly clear. 85 degrees what?
I also think when you're chaining these two separate calculations together you get a problem when it comes to displaying the results.
So yeah, Apple is way behind on this stuff.
Err, what? As a native English speaker human that's a pretty confusing question to me, too!
"As of 2022, there were about 400 million native speakers of English. Including people who speak English as a second language, estimates of the total number of Anglophones vary from 1.5 billion to 2 billion."
Second, all popular models I tested did well with that query, including Gemini on Android (aka "ok Google"), except Apple's.
Some modern Apple devices support "local Siri", but it's a limited subset of both voice recognition performance and capabilities.
Siri needs to be taken out back and shot. The problem with “upgrading” it is the pull to maintain backwards compatibility for every little thing Siri did, which leads them to try and incorporate existing Siri functionality (and existing Siri engineers) to work alongside any LLM. Which leads to disaster, and none of it works and just made it all slower. They’ve been trying to do an LLM assisted Siri for years now and it’s the most public facing disaster the company has had in a while. Time to start over.
Build a crude router in front of it, if you must, or give it access to "the old Siri" as a tool it can call, and let the LLM decide whether to return its own or a Siri-generated response!
I bet even smaller LLMs would be able to figure out, given a user input and Siri response pair, whether the request was resonably answered or whether the model itself could do better or at least explain that the request is out of capabilities for now.
Both of these approaches were tried internally, including even the ability for the LLM to rewrite siri-as-a-tool's response, and none of them shipped, because they all suck. Putting a router in front of it makes multi-turn conversation (when Siri asks for confirmation or disambiguation) a nightmare to implement, and siri-as-a-tool suffers from the same problem. What happens when legacy siri disambiguates? Does the LLM try to guess at an option? Does it proxy the prompt back to the user? What about all the "smart UI" like having a countdown timer with Siri saying "I'll send this" when sending a text message? Does that just pass through? When does the LLM know how/when to intervene in the responses the Siri tool is giving?
This was all an integration nightmare and it's the main reason why none of it shipped. (Well, that and the LLM being underwhelming and the on-device models not being smart enough in the first place. It was just a slower, buggier siri without any new features.)
The answer is that they need to renege on the entire promise of a "private" siri and admit that the only way they can get the experience they want is a _huge_ LLM running with a _ton_ of user context, in the cloud, and don't hinder it all with backwards compatibility with Siri. Give it a toolbox of things it can do with MCP to your device, bake in the stock tools with LoRA or whatever, and let it figure out the best user experience. If it's a frontier-quality LLM it'll be better than Siri on day one, without Apple having to really do anything other than figure out a good system prompt.
The problem is, Apple doesn't want to admit the whole privacy story is a dead-end, so they're going to keep trying to pursue on-device models, and it's going to continue to be underwhelming and "not meeting our quality bar", for the foreseeable future.
But regarding Apple not wanting to admit that client side compute isn't enough: Haven't they essentially already done that, with Private Cloud Computing and all that? I believe not even proofreading and Safari summarization work fully on-device, at least according to my private compute privacy logs.
Yes, but isn't that infuriating? The technology exits! It even exists, as evidenced by this article, in the same company that provides Siri!
At least I feel that way every time I interact with it – or for that matter my Google Home speaker, ironically made and operated by the company that invented transformer networks.
I’m not even sure why those two things would be asked as a single question. It seems like a very unnatural way to pose those two questions. Most humans would trip on that, especially if it was asked verbally.
I'd assume GP only gave an example. As a pretty frequent user, I can unfortunately only confirm that Siri trips over almost every multi-part question.
This would be forgivable if there weren't multiple voice-based AI consumer products available that can handle these kinds of requests perfectly.
If they wanted an LLM answer they could have got one. They went out of their way just to take shots at Apple.
Besides that, many people don’t install any apps, and Apple not pre-installing a reasonable LLM to cater to that market just seems incredibly out of character.
And there’s enough credible reporting and personnel reshuffling happening to suggest that it’s not available yet because they failed to make it work, not because they didn’t try.
(Well, with multiple direct objects, anyway.)
While Siri only does one thing at a time, I trust the answer more, because it’s doing the actual math and not just guessing what the most likely answer is, like an LLM. We need to pick the right tool for the right job. Frankly, I don’t think an LLM is the right tool for conversations like this, and jumbling multiple questions into a single question is something people do with LLMs to get more use out of them during the day, this is an adaptation to a limitation of the free tier (and sometimes speed) of the LLM.
Amazon already reworked Alexa to be backed by a LLM months ago, and they were delayed doing it.
You’re telling me that Apple isn’t capable of the same to Siri?
Amazon just needs a generic LLM. Apple, from the sound of it, is trying to create deep integration with the OS and on-device data. That’s a different problem to solve. They also seem to be trying to do it while respecting user privacy, which is someone most other companies ignore.
I don’t see what the big deal is. I’d rather wait for something good than have them rush out a half-ass “me too” chatbot, that is indistinguishable from the dozens of other chatbots I can simply download as an app for that.
If we believe what Craig Federighi said, they had something, it just wasn’t up to their standards when talking about rolling it out to a billion devices. Which is fair, I run into bad data from ChatGPT and other LLMs all the time. Letting it mature a little more is not a bad thing.
ChatGPT spent a couple months getting my dad pumped up for an elective open heart surgery; he was almost arrogant going into it about how the recovery would go, thinking ChatGPT gave him all the info he could possibly need and a balanced view of reality. Reality hit him pretty hard in the ICU. He sent me some of the chats he had, it was a lot of mutual ego stroking. He was using ChatGPT to downplay the negatives from the doctors and boosting the positives. While it’s good to feel confident, I think it went too far. I spent the whole week in the hospital trying to pull him out of his depression and recalibrating the unrealistic expectations that ChatGPT reinforced. I hope Apple finds a way to be more responsible. If that takes time, great.
Are all Android users using Gemini exclusively? Are all Windows users only using Copilot? Where is the native Linux desktop LLM?
I really don’t understand this criticism. Would it be nice if Siri could do more, sure. Do I have tolerance for Siri to start hallucinating on simple problems it used to use real math for, no. Do I have other options to use in the meantime to get the best of both worlds, absolutely. Where is the hardship?
Much like with the internet, Apple didn't need to re-invent every website to own it all. From Apple platforms a user can access Amazon, Google, or whatever else. Apple didn't create the internet, they sold a gateway to it. AI could be done largely the same way. This way it doesn't matter who wins, Apple can support it. At the end of the day, an LLM doesn't exist on its own, it needs to be accessed through hardware/software people enjoy using, and not be yet another device to charge and carry. Apple has a very popular phone and the most popular wearable. This positions them very well. They are often late to the party, but tend to be best dressed. The first iPhone didn't even have video, and people clowned them for it, and now iPhone video is largely considered one of the best in the smartphone world.
Well partially generated content streaming thing is great and I haven't seen it anywhere else.
(I'm pretty sure this is actually what drove Microsoft Sydney insane.)
Reasoning models can do better at this, because they can write out a good freeform output and then do another pass to transform it.
I'm thinking about let it output freeform and then use another model to use to force that into structured.
By the time Apple has an AI-native product ready, people will already associate it with dehumanization and fascism.
Siri, sideloading and AI features all all the same way; give people options and nobody will complain.
Sideloading is bad for business. Most users don't care. Remember, we, the devs, are not the core target/biggest spenders. They are targeting a large audience of young people who are not tech-savvy.
Why give this to developers if you haven’t been able to get Siri to use it yet? Does it not work or something? I guess we’ll find out when devs start trying to make stuff
Probably Apple is trying to distill the models so they can run on your phone locally. Remember, most, if not all, of Siri is running on your device. There's no round trip whatsoever for voice processing.
Also, for larger models, there will be throwaway VMs per request, so building that infra takes time.
That said, "Private Cloud Compute" does run on proprietary Apple hardware, so availability might be a concern (assuming they don't want to start charging for it).
What exactly are you referring to? Models do run on iPhone and there are features that take advantage of it, today.
The AI stuff with photography sure, but that’s more like machine learning.
The photo touch up thing is… useable? Sometimes?
What is it you’ve been so impressed with?
‘Writing tools’ is there in the text pop up now instead of ‘look up’ even if you’re selecting text on a webpage. It’s just in the way and useless.
(Since it's meant to restrict your children, using it to restrict yourself will disable some features that'd let you escape it. I forget what exactly, but you might not be able to change the time or something like that.)
> Ok its doing it in 4 or 5 products, but thats a joke.
Not every AI product is a chatbot.
Wow.
I'm currently on the beta, and I have a shortcut that pulls in various bits of context and feeds that directly to the on-device model.
I mean, the thing even lets me ask ChatGPT things if I explicitly ask it to! But why do I need to ask in the first place?
I don’t speak for Apple but certainly you can appreciate that there is a fine balance between providing basic functionality and providing apps. Apple works to provide tools that developers can leverage and also tries to not step on those same developers. Defining the line between what should be built-in and what should be add-on needs to be done carefully and often is done organically.
Q: Should search be core behavior or third-party functionality?
My son (he's 11 years old now and fairly skilled with all the main AI tools, eg chatgpt, gemini, etc) and I retry her every month or so, and this past time we just laughed. Can't handle basic questions - hears the question wrong, starts, stops, takes us to some random ass webpage, etc, etc.
"She's so jacked up!" he said.
Apple needs to get this under control and figured out, stat!
Running on device is also risky because cycle limitations will make it seem dumb in comparison.
Tim Cook might be better at squeezing the juice, but he is not a product person.
This time around they need another solution, otherwise regardless of how much money they have, they will stay as the iOS/iPad company, given the relevance of macOS on desktop market worldwide.
https://share.icloud.com/photos/018AYAPEm06ALXciiJAsLGyuA
https://share.icloud.com/photos/0f9IzuYQwmhLIcUIhIuDiudFw
The above took like 3 seconds to generate. That little box that says On-device can be flipped between On-device, Private Cloud Compute, and ChatGPT.
Their LLM uses the ANE sipping battery and leaves the GPU available.
There's a WWDC video "Meet the Foundation Models Framework" [1].
And this is after they made very big claims with Apple Intelligence last year, when they had everyone fooled.
This is like watching a train-wreck in slow motion.
Yes, this is in fact what people want. Apple is the biggest company in the world (don’t quibble this y’all, you know what I mean) and should be able to deliver this experience. And sure, if they could do it on device that would be aces, but that’s not an item on the menu, and customers seem fine with web-based things like ChatGPT for now. To act like Apple is doing anything other than fumbling right now is cope.
The people running large models want to charge a monthly fee for that.
I'm fine with having a free model that runs on device without slurping up my data.
It's hard to be like "uhhh privacy" when you send all requests to a remote server where they're stored in clear text for god knows how long.
As of right now, there is no way to run big LLMs in a privacy preserving manner. It just doesn't exist. You can't E2EE encrypt these services, because the compute is done on the server, so it has to decrypt it.
There are some services which will randomize your instance and things like that, but that kind of defeats the a big part of what makes LLMs useful, context. Until we can run these models locally, there's no way to get around the privacy nightmare aspects of it.
Siri, even after decades of investment, is a joke. Apple does NOT have the talent or capability to deliver what people want.
I don’t really understand why Apple has to provide a ChatGPT product, baked directly into their software. Why on earth would Apple want to get involved in the race to the bottom for the cheapest LLMs? Apple doesn’t produce commodity products, they package commodities into something much more unique that gives them a real competitive advantage, so people are willing to pay a premium for the Apple’s product, rather than just buying the cheapest commodity equivalent.
There is no point Apple just delivering an LLM. OpenAI, Anthropic, Google etc already do that, and Apple is never going to get into the pay-per-call API service they all offer. Delivering AI experiences using on-device only compute, that’s something OpenAI, Anthropic and Google can’t build, which means Apple can easily charge an premium for it, assuming they build it.
Control. It boils down to control. If you own a platform, you want to make your "suppliers" (apps in this case) as substitutable as possible.
If people start associating ChatGPT or Claude or Gemini as the main reasons to buy a phone, at some point in the future, they'll think - gee, most of what I'm doing on the phone is interacting with $app, and I can get the $app elsewhere.
With my history encrypted in the cloud, and the trust that Apple has built around privacy ... I think they're going to come out alright.
This is the first time in 10+ years I've seen Apple so far on the back foot. They usually launch category defining products that are so far ahead of the competition, even by the time they work through the 'drawbacks' in the first versions of them they are still far ahead. OS X, the iPhone and the iPad were all like that. They are still way ahead of the competition on Apple Silicon as well.
I am not very confident on their on device strategy at least in the short to medium term. Nearly all their devices do not have enough RAM and even if they did SLMs are very far behind what users "know" as AI - even the free ChatGPT plan is leap years ahead of the best 3B param on device model. Maybe there will be huge efficiency gains.
Private cloud is used AFIAK for virtually 0 use cases so far. Perhaps it will be more interesting longer term but not very useful at the moment given the lack of a suitable (ie: non Chinese), large (>500b param) model. They would also struggle to scale it if they roll it out to billions of iOS devices especially if they put features that use a lot of tokens.
Then they've got OpenAI/Gemini/Anthropic via API. But this completely goes against all their private cloud messaging and gives those providers enormous potential control over Apple, which is not a position Apple usually finds itself in. It will also be extremely expensive to pay someone per token for OS level features for billions of iOS/Mac devices and unless they can recoup this via some sort of subscription will hit services margins badly.
To me its clear the future of "OS" is going to involve a lot of agentic tool calling. These require good models, with large context windows and a lot of tokens - this will definitely not work on device. Indeed this is exactly what the Siri vapourware demo was.
I'm sure they can potentially get to a great UX (though these missteps are making me question this). But having such a core feature outsourced does not leave them in a good position.
Applications using Apple's foundation models can seamlessly switch from on-device models to Private Compute Cloud.
Research is already showing the use of LLMs for people's most intimate relationship and medical issues. The usual suspects will try to monetize that, which why Private Cloud Compute is a thing from the jump.
> Then they've got OpenAI/Gemini/Anthropic via API. But this completely goes against all their private cloud messaging
Using ChatGPT via Siri today, no personally identifying information is shared with OpenAI and those prompts aren't used for training. I suspect Apple would want something similar for Google, Anthropic, etc.
At some point, there will be the inevitable enshitification of AI platforms to recoup the billions VCs have invested, which means ads, which won't happen to Apple users using foundation model-based apps.
> Nearly all their devices do not have enough RAM and
Every Apple Silicon Mac (going back to the M1 in 2020) can run Apple Intelligence. 8 GB RAM is all they need. Every iPhone 15 Pro, Pro Max and the entire 16 line can all run Apple Intelligence.
Flagship iPhone 17 models are expected to come with 12 GB of RAM and all current Mac models come with at least 16 GB.
Apple sells over 200 million iPhones in a given year.
There's no doubt Apple stumbled out of the gate regarding AI; these are early days. They can't be counted out.
Even still though people are used to the quality of huge frontier models, so it will feel like a massive downgrade on many tasks. The _big_ problem with all this is chained tool calling. It uses context SO quickly and context needs a lot of (V)RAM. This also completely undermines the privacy argument you make, because it will need to ask personal data if using OpenAI and put it in the prompt.
Yes I noticed Apple shipping higher RAM but it will take years for this to feed through to a sizeable userbase, and people are quickly getting ingrained to use an app like ChatGPT instead of OS level features. Even more so given what a flop Apple Intelligence 1.0 has been.
The key problem they've got is they've went hard on privacy (which means it is hard to square that with going all in on 3rd party APIs) but they've also been incredibly stingy with RAM historically, which really nerfs their on device options. Private compute is an interesting middle ground but their model options are incredibly limited currently.
Apple's ~3 billion parameter on-device model is about as good as it gets on a smartphone, especially for the functions it was designed for: writing and refining text, prioritizing and summarizing notifications, creating images for conversations, and taking in-app actions.
Every Mac comes with at least 16 GB of RAM; while every iPhone comes with 8 GB of RAM, some models of the iPhone 17 will have 12 GB.
Remember, an app using the on-device model can seamlessly shift to a much bigger model via Private Cloud Compute without the user having to do anything.
If the user enables it, Apple's Foundation Model can use ChatGPT in a privacy preserving way. By the fall, Gemini and Sonnet/Opus could be options as well.
Again, ChatGPT is used in a privacy preserving way; you don't need an account: "Use ChatGPT with Apple Intelligence on iPhone" [1].
[1]: https://support.apple.com/guide/iphone/use-chatgpt-with-appl...
It's already accessible using Shortcuts, even to non-developers "iOS 26 Shortcuts + Apple Intelligence is POWERFUL " (Youtube) [1].
They were touting the same features that other companies are now delivering. Point the phone at something, and it'll tell you what you're looking at. Or summarize news articles etc. Instead we got .. emojithingy
They're not a model company. The risks of deploying something half-baked to their users is unacceptable. They're taking it slow and trying to do it in a way that doesn't damage/erode their brand.
Wait it out, let the best model(s) rise to the surface (and the hallucination problems to get sufficiently mitigated), and then either partner with a proprietary provider or deploy one of the open source models. Makes more sense than burning billions of dollars training a new foundation model
https://www.theverge.com/2024/12/13/24320689/apple-intellige...
Apple urged to withdraw 'out of control' AI news alerts
https://www.bbc.com/news/articles/cge93de21n0o
iOS 18.3 Temporarily Removes Notification Summaries for News
https://www.reddit.com/r/apple/comments/1i2w65j/ios_183_temp...
They love to "invent" brands that they control, so that they can commodotize the underlying supplier. Hey user, it is a retina display and dont worry whether it is LG or Samsung is making it.
Apple tried this with AI, calling it "Apple Intelligence". Unfortunately that faltered. Now Apple will have to come out and say "iPhone with ChatGPT" or "Siri with Claude". AND APPLE HATES THAT. HATES IT WITH PASSION.
People will start to associate smartness with ChatGPT or Claude, and Apple loses control and OpenAI/Anthropic's leverage goes up.
Apple has painted themselves into a corner. And as I said elsewhere, it is a train-wreck happening in slowmotion.
Or consider that they spent a decade highlighting that their computers were powered by Intel, after leaving their proprietary PowerPC architecture—again, under Steve Jobs.
Or go all the way back to 1997 when Steve Jobs had Bill Gates on the screen at Macworld and announced that IE would be the default browser on Mac.
It’s easy to fall into a caricature of Apple, where they insist on making everything themselves. What is more accurate is to say that they are not afraid to make things themselves, when they think they have a better idea. But they are also not afraid to do deals when it is the best way forward right now.
What is this train-wreck you are hallucinating?
Most of his team are former Google brain so GDM knows who is good.
This is the first time that millions of people will actually download and run a model on their own devices.
The question is… will Apple be constantly tweaking these models, or only during OS upgrades?
I for one really like local software. Call me old-fashioned, but I enjoy when a company doesn’t switch up software anytime on the server, or phone the results home all the time in order to extract more profits from their collective users.
Certainly when new updates are released--going from macOS 26 to 26.1).
They can probably push model updates between releases if necessary.
> “Adapters produced by the toolkit are fully compatible with the Foundation Models framework. However, each adapter is compatible with a single specific model version, meaning that a new adapter must be trained for each new version of the base model.”
Any changes should require retraining any LoRA adapters that has been built & distributed by third party developers, so they wouldn’t update the models outside OS updates at the drop of a hat I don’t think.
LoRA adapters can be distributed via Background Assets, but the base model itself should be version-locked to the OS build (e.g. iOS 26.0 → 26.1) and updates only when Apple ships a new OS image.