> For those just getting started, developers who are enrolled in the App Store Small Business Program and have fewer than two million first-time App Store downloads can access the Apple Foundation Models running on Private Cloud Compute with no cloud API cost.*
> Where Apple Intelligence is available, eligible developers can use PCC in their apps distributed on the App Store, and test PCC features via TestFlight or ad hoc distribution. Installs during testing are not counted as first-time app downloads.
Is this free tokens? This sounds like free tokens?
Pretty much. You need an Organization developer account, not an Individual account to get the PCC entitlement, but after that it’s pretty easy. I’ve been using it in TestFlight and now a shipping app since WWDC and it’s pretty solid. A lot more performant than I had expected. Context windows could be a bit larger (though it’s much better than the on-device model). PCC usage is allotted to the end users Apple ID or iCloud account , not the app or my development account, which makes it possible to ship a lightweight app and not worry about my own server costs. Some of the background gotchas and PCC rate limiting behaviors are obnoxious but I’ll take what I can get. Some of it is just standard iOS behavior.
Yep. Though "free tokens" with unspecified daily limits is doing a lot of heavy lifting here. Still, for a small indie app that only needs occasional inference, this could be a pretty big deal
Not exactly. They’re not your tokens, they’re the user’s tokens. If those are different people.
Your app uses tokens when it goes to private cloud compute tied to the user’s iCloud account. I don’t think anyone knows what the limit is, we only know that higher iCloud plans bring higher limits.
It’s all very fuzzy right now as this is brand new and I think they’re just trying to incentivize development rather than risk strangling it out of the gate.
Interesting. Have you had issues with the model safety restrictions at all? They seem rather aggressive for normal usage, I wonder if limiting to command line options keeps things safe enough they stay out of the way.
I have my firewall events get reviewed by the local model on my mac studio. I had to stop sending the full event because the model would refuse some requests.
I actually think their CoreAI stuff is kind of flying under the radar. But they've got a repo at https://github.com/apple/coreai-models where they're adapting a ton of open weight models to their format to run natively with Swift.
I just realized that claude has also support App Attest for months [1]. Not sure if that is being used widely. Firebase was the most practicle option to call llms directly from ios apps without having fancy setups. I am surprised Anthropic not promoting this more.
This was the first time I've encountered App Attest and it seems like a pretty solid solution to the problem of having to trust that a request from someone else's device is in fact originating from your own app rather than an extracted key of some form.
I have been using Foundation Models in Noodle for a while and have liked working with it. One part I have not explored yet is Apple Private Cloud Compute but it is on my list. I think I need special entitlements that are only available to Siri desktop app for now - need to check.
I just had a very basic poke at `fm` which runs models locally it was at least able to spit out fizzbuzz in both Python and Rust impressively fast. Whether its any good wrapped in a coding harness remains to be seen.
Even the stronger Private Cloud Compute model struggled with the coding task I gave it (build a web-based side-scrolling platformer). I used this and pointed Cursor at it to test: https://github.com/littlebobert/pcc-server-example
macOS 27 comes with the pre-installed fm cli command, which is a harness for Apple's local models. Local models run on-device without limits. It's a nice tool to play with — although it does have heavy guardrails.
> The Foundation Models framework now extends to server models — like Claude, Gemini, and more — all through the same Swift API, with full support for features like tool calling and guided generation
Claude, Gemini, and more in eh? Seems like a big one is missing from that explicit list haha.
Not a big surprise ChatGPT isn’t directly mentioned given their legal spat, just kind of funny to see the pettiness.
We've reached the point where Apple is offering a unified abstraction over competing AI providers. Somewhere, a developer is already figuring out which provider-specific feature breaks the abstraction first.
> For those just getting started, developers who are enrolled in the App Store Small Business Program and have fewer than two million first-time App Store downloads can access the Apple Foundation Models running on Private Cloud Compute with no cloud API cost.*
https://developer.apple.com/private-cloud-compute/
> Where Apple Intelligence is available, eligible developers can use PCC in their apps distributed on the App Store, and test PCC features via TestFlight or ad hoc distribution. Installs during testing are not counted as first-time app downloads.
Is this free tokens? This sounds like free tokens?
Pretty much. You need an Organization developer account, not an Individual account to get the PCC entitlement, but after that it’s pretty easy. I’ve been using it in TestFlight and now a shipping app since WWDC and it’s pretty solid. A lot more performant than I had expected. Context windows could be a bit larger (though it’s much better than the on-device model). PCC usage is allotted to the end users Apple ID or iCloud account , not the app or my development account, which makes it possible to ship a lightweight app and not worry about my own server costs. Some of the background gotchas and PCC rate limiting behaviors are obnoxious but I’ll take what I can get. Some of it is just standard iOS behavior.
You can have it with an individual developer account (as long as it's a paid developer account). I have it in mine.
Ah you are correct, I was misremembering the timing of some other account changes I had to make and associating them with this.
Yep. Though "free tokens" with unspecified daily limits is doing a lot of heavy lifting here. Still, for a small indie app that only needs occasional inference, this could be a pretty big deal
This sounds like free samples of tokens.
Not exactly. They’re not your tokens, they’re the user’s tokens. If those are different people.
Your app uses tokens when it goes to private cloud compute tied to the user’s iCloud account. I don’t think anyone knows what the limit is, we only know that higher iCloud plans bring higher limits.
It’s all very fuzzy right now as this is brand new and I think they’re just trying to incentivize development rather than risk strangling it out of the gate.
I’ve been using these in my terminal for correcting errors and suggesting commands https://github.com/paul-oms/prompt
Surprisingly capable on an m1 8gb (in my opinion this is the smallest machine that local ai needs to run on)
Interesting. Have you had issues with the model safety restrictions at all? They seem rather aggressive for normal usage, I wonder if limiting to command line options keeps things safe enough they stay out of the way.
Yeah I found the safety restrictions really aggressive. Enough that I just gave up and used Gemma instead.
I have my firewall events get reviewed by the local model on my mac studio. I had to stop sending the full event because the model would refuse some requests.
I actually think their CoreAI stuff is kind of flying under the radar. But they've got a repo at https://github.com/apple/coreai-models where they're adapting a ton of open weight models to their format to run natively with Swift.
I just realized that claude has also support App Attest for months [1]. Not sure if that is being used widely. Firebase was the most practicle option to call llms directly from ios apps without having fancy setups. I am surprised Anthropic not promoting this more.
[1] https://github.com/anthropics/ClaudeForFoundationModels#auth...
This was the first time I've encountered App Attest and it seems like a pretty solid solution to the problem of having to trust that a request from someone else's device is in fact originating from your own app rather than an extracted key of some form.
Why is this up front on HN? Did it get an update?
Discussed last year when it was announced https://news.ycombinator.com/item?id=44226978
There have been a few published rumors of Apple getting ready to release a new set of local models.
Where did you see these rumors?
I have been using Foundation Models in Noodle for a while and have liked working with it. One part I have not explored yet is Apple Private Cloud Compute but it is on my list. I think I need special entitlements that are only available to Siri desktop app for now - need to check.
This is the repo if you are interested https://github.com/pdparchitect/noodle
Slightly off topic, but I haven't seen benchmarks or usage tests of the on-device model against other similarly sized models.
I know you can only expect so much from a 3B model, but how does it compare to e.g. Gemma 4 E4B or Qwen 3.5 4B?
Can the local model code at all? Eg as the predictor in Zed?
I just had a very basic poke at `fm` which runs models locally it was at least able to spit out fizzbuzz in both Python and Rust impressively fast. Whether its any good wrapped in a coding harness remains to be seen.
Even the stronger Private Cloud Compute model struggled with the coding task I gave it (build a web-based side-scrolling platformer). I used this and pointed Cursor at it to test: https://github.com/littlebobert/pcc-server-example
Curious if it’s possible to use a non-API plan for Claude/ChatGPT with this? It could actually be pretty useful.
macOS 27 comes with the pre-installed fm cli command, which is a harness for Apple's local models. Local models run on-device without limits. It's a nice tool to play with — although it does have heavy guardrails.
Under the “Model Support” section:
> The Foundation Models framework now extends to server models — like Claude, Gemini, and more — all through the same Swift API, with full support for features like tool calling and guided generation
Claude, Gemini, and more in eh? Seems like a big one is missing from that explicit list haha.
Not a big surprise ChatGPT isn’t directly mentioned given their legal spat, just kind of funny to see the pettiness.
We've reached the point where Apple is offering a unified abstraction over competing AI providers. Somewhere, a developer is already figuring out which provider-specific feature breaks the abstraction first.
Siri AI is too clever!!!!!
Siri AI is very clever
Are we talking about the same Siri? Because mine still occasionally struggles to set a timer
I'll ask my LLM to build me a framework, thank you.
Why would you do that, assuming that you're not coding yourself/using the framework?