At WWDC25 we launched a new type of Lab event for the developer community - Group Labs. A Group Lab is a panel Q&A designed for a large audience of developers. Group Labs are a unique opportunity for the community to submit questions directly to a panel of Apple engineers and designers. Here are the highlights from the WWDC25 Group Lab for Machine Learning and AI Frameworks. What are you most excited about in the Foundation Models framework? The Foundation Models framework provides access to an on-device Large Language Model (LLM), enabling entirely on-device processing for intelligent features. This allows you to build features such as personalized search suggestions and dynamic NPC generation in games. The combination of guided generation and streaming capabilities is particularly exciting for creating delightful animations and features with reliable output. The seamless integration with SwiftUI and the new design material Liquid Glass is also a major advantage. When should I still bring my own LLM via CoreML? It's generally recommended to first explore Apple's built-in system models and APIs, including the Foundation Models framework, as they are highly optimized for Apple devices and cover a wide range of use cases. However, Core ML is still valuable if you need more control or choice over the specific model being deployed, such as customizing existing system models or augmenting prompts. Core ML provides the tools to get these models on-device, but you are responsible for model distribution and updates. Should I migrate PyTorch code to MLX? MLX is an open-source, general-purpose machine learning framework designed for Apple Silicon from the ground up. It offers a familiar API, similar to PyTorch, and supports C, C++, Python, and Swift. MLX emphasizes unified memory, a key feature of Apple Silicon hardware, which can improve performance. It's recommended to try MLX and see if its programming model and features better suit your application's needs. MLX shines when working with state-of-the-art, larger models. Can I test Foundation Models in Xcode simulator or device? Yes, you can use the Xcode simulator to test Foundation Models use cases. However, your Mac must be running macOS Tahoe. You can test on a physical iPhone running iOS 18 by connecting it to your Mac and running Playgrounds or live previews directly on the device. Which on-device models will be supported? any open source models? The Foundation Models framework currently supports Apple's first-party models only. This allows for platform-wide optimizations, improving battery life and reducing latency. While Core ML can be used to integrate open-source models, it's generally recommended to first explore the built-in system models and APIs provided by Apple, including those in the Vision, Natural Language, and Speech frameworks, as they are highly optimized for Apple devices. For frontier models, MLX can run very large models. How often will the Foundational Model be updated? How do we test for stability when the model is updated? The Foundation Model will be updated in sync with operating system updates. You can test your app against new model versions during the beta period by downloading the beta OS and running your app. It is highly recommended to create an "eval set" of golden prompts and responses to evaluate the performance of your features as the model changes or as you tweak your prompts. Report any unsatisfactory or satisfactory cases using Feedback Assistant. Which on-device model/API can I use to extract text data from images such as: nutrition labels, ingredient lists, cashier receipts, etc? Thank you. The Vision framework offers the RecognizeDocumentRequest which is specifically designed for these use cases. It not only recognizes text in images but also provides the structure of the document, such as rows in a receipt or the layout of a nutrition label. It can also identify data like phone numbers, addresses, and prices. What is the context window for the model? What are max tokens in and max tokens out? The context window for the Foundation Model is 4,096 tokens. The split between input and output tokens is flexible. For example, if you input 4,000 tokens, you'll have 96 tokens remaining for the output. The API takes in text, converting it to tokens under the hood. When estimating token count, a good rule of thumb is 3-4 characters per token for languages like English, and 1 character per token for languages like Japanese or Chinese. Handle potential errors gracefully by asking for shorter prompts or starting a new session if the token limit is exceeded. Is there a rate limit for Foundation Models API that is limited by power or temperature condition on the iPhone? Yes, there are rate limits, particularly when your app is in the background. A budget is allocated for background app usage, but exceeding it will result in rate-limiting errors. In the foreground, there is no rate limit unless the device is under heavy load (e.g., camera open, game mode). The system dynamically balances performance, battery life, and thermal conditions, which can affect the token throughput. Use appropriate quality of service settings for your tasks (e.g., background priority for background work) to help the system manage resources effectively. Do the foundation models support languages other than English? Yes, the on-device Foundation Model is multilingual and supports all languages supported by Apple Intelligence. To get the model to output in a specific language, prompt it with instructions indicating the user's preferred language using the locale API (e.g., "The user's preferred language is en-US"). Putting the instructions in English, but then putting the user prompt in the desired output language is a recommended practice. Are larger server-based models available through Foundation Models? No, the Foundation Models API currently only provides access to the on-device Large Language Model at the core of Apple Intelligence. It does not support server-side models. On-device models are preferred for privacy and for performance reasons. Is it possible to run Retrieval-Augmented Generation (RAG) using the Foundation Models framework? Yes, it is possible to run RAG on-device, but the Foundation Models framework does not include a built-in embedding model. You'll need to use a separate database to store vectors and implement nearest neighbor or cosine distance searches. The Natural Language framework offers simple word and sentence embeddings that can be used. Consider using a combination of Foundation Models and Core ML, using Core ML for your embedding model.

Hi From https://developer.apple.com/metal/jax/ I checked all active workflows on https://github.com/jax-ml/jax and any open issues with tags Metal and seems in DEC 2025 the Jax maintainers have closed all issues citing No active development on Jax-metal and the project seems dead. We need to know how can we leverage Apple silicon for accelerated projects using popular academia library and tools . Is the JAX project still going to be supported or Apple has plans to bring something of tis own that might be platform agnostic . Thanks

Hardware: Macbook Pro M4 Nov 2024 Software: macOS Tahoe 26.0 & xcode 26.0 Apple Intelligence is activated and the Image playground macOS app works Running the following on xcode throws ImagePlayground.ImageCreator.Error.creationFailed Any suggestions on how to make this work? import Foundation import ImagePlayground Task { let creator = try await ImageCreator() guard let style = creator.availableStyles.first else { print("No styles available") exit(1) } let images = creator.images( for: [.text("A cat wearing mittens.")], style: style, limit: 1) for try await image in images { print("Generated image: \(image)") } exit(0) } RunLoop.main.run()

I get the following error when running this command in a Jupyter notebook: v = tf.Variable(initial_value=tf.random.normal(shape=(3, 1))) v[0, 0].assign(3.) Environment: python == 3.11.14 tensorflow==2.19.1 tensorflow-metal==1.2.0 { "name": "InvalidArgumentError", "message": "Cannot assign a device for operation ResourceStridedSliceAssign: Could not satisfy explicit device specification '/job:localhost/replica:0/task:0/device:GPU:0' because no supported kernel for GPU devices is available.\nColocation Debug Info:\nColocation group had the following types and supported devices: \nRoot Member(assigned_device_name_index_=1 requested_device_name_='/job:localhost/replica:0/task:0/device:GPU:0' assigned_device_name_='/job:localhost/replica:0/task:0/device:GPU:0' resource_device_name_='/job:localhost/replica:0/task:0/device:GPU:0' supported_device_types_=[CPU] possible_devices_=[]\nResourceStridedSliceAssign: CPU \n_Arg: GPU CPU \n\nColocation members, user-requested devices, and framework assigned devices, if any:\n ref (_Arg) framework assigned device=/job:localhost/replica:0/task:0/device:GPU:0\n ResourceStridedSliceAssign (ResourceStridedSliceAssign) /job:localhost/replica:0/task:0/device:GPU:0\n\nOp: ResourceStridedSliceAssign\n [...] [[{{node ResourceStridedSliceAssign}}]] [Op:ResourceStridedSliceAssign] name: strided_slice/_assign" } It seems like the ResourceStridedSliceAssign operation is not implemented for the GPU

Posting a follow up question after the WWDC 2025 Machine Learning AI & Frameworks Group Lab on June 12. In regards to the on-device API of any of the AI frameworks (foundation model, vision framework, ect.), is there a response condition or path where the API outsources it's input to ChatGPT if the user has allowed this like Siri does? Ignore this if it's a no: is this handled behind the scenes or by the developer?

I'm using Vision framework (DetectFaceLandmarksRequest) with the same code and the same test image to detect face landmarks. On iOS 18 everything works as expected: detected face landmarks align with the face correctly. But when I run the same code on devices with iOS 26, the landmark coordinates are outside the [0,1] range, which indicates they are out of face bounds. Fun fact: the old VNDetectFaceLandmarksRequest API works very well without encountering this issue How I get face landmarks: private let faceRectangleRequest = DetectFaceRectanglesRequest(.revision3) private var faceLandmarksRequest = DetectFaceLandmarksRequest(.revision3) func detectFaces(in ciImage: CIImage) async throws -> FaceTrackingResult { let faces = try await faceRectangleRequest.perform(on: ciImage) faceLandmarksRequest.inputFaceObservations = faces let landmarksResults = try await faceLandmarksRequest.perform(on: ciImage) ... } How I show face landmarks in SwiftUI View: private func convert( point: NormalizedPoint, faceBoundingBox: NormalizedRect, imageSize: CGSize ) -> CGPoint { let point = point.toImageCoordinates( from: faceBoundingBox, imageSize: imageSize, origin: .upperLeft ) return point } At the same time, it works as expected and gives me the correct results: region is FaceObservation.Landmarks2D.Region let points: [CGPoint] = region.pointsInImageCoordinates( imageSize, origin: .upperLeft ) After that, I found that the landmarks are normalized relative to the unalignedBoundingBox. However, I can’t access it in code. Still, using these values for the bounding box works correctly. Things I've already tried: Same image input Tested multiple devices on iOS 26.2 -> always wrong. Tested multiple devices on iOS 18.7.1 -> always correct. Environment: macOS 26.2 Xcode 26.2 (17C52) Real devices, not simulator Face Landmarks iOS 18 Face Landmarks iOS 26

I am writing a custom package wrapping Foundation Models which provides a chain-of-thought with intermittent self-evaluation among other things. At first I was designing this package with the command line in mind, but after seeing how well it augments the models and makes them more intelligent I wanted to try and build a SwiftUI wrapper around the package. When I started I was using synchronous generation rather than streaming, but to give the best user experience (as I've seen in the WWDC sessions) it is necessary to provide constant feedback to the user that something is happening. I have created a super simplified example of my setup so it's easier to understand. First, there is the Reasoning conversation item, which can be converted to an XML representation which is then fed back into the model (I've found XML works best for structured input) public typealias ConversationContext = XMLDocument extension ConversationContext { public func toPlainText() -> String { return xmlString(options: [.nodePrettyPrint]) } } /// Represents a reasoning item in a conversation, which includes a title and reasoning content. /// Reasoning items are used to provide detailed explanations or justifications for certain decisions or responses within a conversation. @Generable(description: "A reasoning item in a conversation, containing content and a title.") struct ConversationReasoningItem: ConversationItem { @Guide(description: "The content of the reasoning item, which is your thinking process or explanation") public var reasoningContent: String @Guide(description: "A short summary of the reasoning content, digestible in an interface.") public var title: String @Guide(description: "Indicates whether reasoning is complete") public var done: Bool } extension ConversationReasoningItem: ConversationContextProvider { public func toContext() -> ConversationContext { // <ReasoningItem title="${title}"> // ${reasoningContent} // </ReasoningItem> let root = XMLElement(name: "ReasoningItem") root.addAttribute(XMLNode.attribute(withName: "title", stringValue: title) as! XMLNode) root.stringValue = reasoningContent return ConversationContext(rootElement: root) } } Then there is the generator, which creates a reasoning item from a user query and previously generated items: struct ReasoningItemGenerator { var instructions: String { """ <omitted for brevity> """ } func generate(from input: (String, [ConversationReasoningItem])) async throws -> sending LanguageModelSession.ResponseStream<ConversationReasoningItem> { let session = LanguageModelSession(instructions: instructions) // build the context for the reasoning item out of the user's query and the previous reasoning items let userQuery = "User's query: \(input.0)" let reasoningItemsText = input.1.map { $0.toContext().toPlainText() }.joined(separator: "\n") let context = userQuery + "\n" + reasoningItemsText let reasoningItemResponse = try await session.streamResponse( to: context, generating: ConversationReasoningItem.self) return reasoningItemResponse } } I'm not sure if returning LanguageModelSession.ResponseStream<ConversationReasoningItem> is the right move, I am just trying to imitate what session.streamResponse returns. Then there is the orchestrator, which I can't figure out. It receives the streamed ConversationReasoningItems from the Generator and is responsible for streaming those to SwiftUI later and also for evaluating each reasoning item after it is complete to see if it needs to be regenerated (to keep the model on-track). I want the users of the orchestrator to receive partially generated reasoning items as they are being generated by the generator. Later, when they finish, if the evaluation passes, the item is kept, but if it fails, the reasoning item should be removed from the stream before a new one is generated. So in-flight reasoning items should be outputted aggresively. I really am having trouble figuring this out so if someone with more knowledge about asynchronous stuff in Swift, or- even better- someone who has worked on the Foundation Models framework could point me in the right direction, that would be awesome!

Can't able to run the Create ML for training and I upgraded to MacOS 26.3 beta and I have tried older and newer

In this WWDC25 session, it is explictely mentioned that apps should support AttributedString for text parameters to their App Intents. However, I have not gotten this to work. Whenever I pass rich text (either generated by the new "Use Model" intent or generated manually for example using "Make Rich Text from Markdown"), my Intent gets an AttributedString with the correct characters, but with all attributes stripped (so in effect just plain text). struct TestIntent: AppIntent { static var title = LocalizedStringResource(stringLiteral: "Test Intent") static var description = IntentDescription("Tests Attributed Strings in Intent Parameters.") @Parameter var text: AttributedString func perform() async throws -> some IntentResult & ReturnsValue<AttributedString> { return .result(value: text) } } Is there anything else I am missing?

hello, Do you have any information on the handling of sparse matrix with MPS and PyTorch? release date? ...

I'm trying the new RecognizeDocumentsRequest supposed to detect paragraphs (among other things) in a document. I tried many source images, and I don't see the slightest difference compared to the old API (VN)RecognizedTextRequest Is it supposed to not work or is it in beta?

I'm implementing an App Intent for my iOS app that helps users plan trip activities. It only works when run as a shortcut but not using voice through Siri. There are 2 issues: The ShortcutsTripEntity will only accept a voice input for a specific trip but not others. I'm stuck with a throwing error when trying to use requestDisambiguation() on the activity day @Parameter property. How do I rectify these issues. This is blocking me from completing a critical feature that lets users quickly plan activities through Siri and Shortcuts. Expected behavior for trip input: The intent should make Siri accept the spoken trip input from any of the options. Actual behavior for trip input: Siri only accepts the same trip when spoken but accepts any when selected by click/touch. Expected behavior for day input: Siri should accept the spoken selected option. Actual behavior for day input: Siri only accepts an input by click/touch but yet throws an error at runtime I'm happy to provide more code. But here's the relevant code: struct PlanActivityTestIntent: AppIntent { @Parameter(title: "Activity Day") var activityDay: ShortcutsItineraryDayEntity @Parameter( title: "Trip", description: "The trip to plan an activity for", default: ShortcutsTripEntity(id: UUID().uuidString, title: "Untitled trip"), requestValueDialog: "Which trip would you like to add an activity to?" ) var tripEntity: ShortcutsTripEntity @Parameter(title: "Activity Title", description: "The title of the activity", requestValueDialog: "What do you want to do or see?") var title: String @Parameter(title: "Activity Day", description: "Activity Day", default: ShortcutsItineraryDayEntity(itineraryDay: .init(itineraryId: UUID(), date: .now), timeZoneIdentifier: "UTC")) var activityDay: ShortcutsItineraryDayEntity func perform() async throws -> some ProvidesDialog { // ...other code... let tripsStore = TripsStore() // load trips and map them to entities try? await tripsStore.getTrips() let tripsAsEntities = tripsStore.trips.map { trip in let id = trip.id ?? UUID() let title = trip.title return ShortcutsTripEntity(id: id.uuidString, title: title, trip: trip) } // Ask user to select a trip. This line would doesn't accept a voice // answer. Why? let selectedTrip = try await $tripEntity.requestDisambiguation( among: tripsAsEntities, dialog: .init( full: "Which of the \(tripsAsEntities.count) trip would you like to add an activity to?", supporting: "Select a trip", systemImageName: "safari.fill" ) ) // This line throws an error let selectedDay = try await $activityDay.requestDisambiguation( among: daysAsEntities, dialog:"Which day would you like to plan an activity for?" ) } } Here are some related images that might help:

In an under-development MacOS & iOS app, I need to identify various measurements from OCR'ed text: length, weight, counts per inch, area, percentage. The unit type (e.g. UnitLength) needs to be identified as well as the measurement's unit (e.g. .inches) in order to convert the measurement to the app's internal standard (e.g. centimetres), the value of which is stored the relevant CoreData entity. The use of NLTagger and NLTokenizer is problematic because of the various representations of the measurements: e.g. "50g.", "50 g", "50 grams", "1 3/4 oz." Currently, I use a bespoke algorithm based on String contains and step-wise evaluation of characters, which is reasonably accurate but requires frequent updating as further representations are detected. I'm aware of the Python SpaCy model being capable of NER Measurement recognition, but am reluctant to incorporate a Python-based solution into a production app. (ref [https://developer.apple.com/forums/thread/30092]) My preference is for an open-source NER Measurement model that can be used as, or converted to, some form of a Swift compatible Machine Learning model. Does anyone know of such a model?

I'm on Tahoe 26.1 / M3 Macbook Air. I'm using VNDetectFaceRectanglesRequest as properly as possible, as in the minimal command line program attached below. For some reason, I always get: MLE5Engine is disabled through the configuration printed. I couldn't find any notes on developer docs saying that VNDetectFaceRectanglesRequest can not use the Apple Neural Engine. I'm assuming there is something wrong with my code however I wasn't able to find any remarks from documentation where it might be. I wasn't able to find the above error message online either. I would appreciate your help a lot and thank you in advance. The code below accesses the video from AVCaptureDevice.DeviceType.builtInWideAngleCamera. Currently it directly chooses the 0th format which has the largest resolution (Full HD on my M3 MBA) and "4:2:0" color "v" reduced color component spectrum encoding ("420v"). After accessing video, it performs a VNDetectFaceRectanglesRequest. It prints "VNDetectFaceRectanglesRequest completion Handler called" many times, then prints the error message above, then continues printing "VNDetectFaceRectanglesRequest completion Handler called" until the user quits it. To run it in Xcode, File > New project > Mac command line tool. Pasting the code below, then click on the root file > Targets > Signing & Capabilities > Hardened Runtime > Resource Access > Camera. A possible explanation could be that either Apple's internal CoreML code for this function works on GPU/CPU only or it doesn't accept 420v as supplied by the Macbook Air camera import AVKit import Vision var videoDataOutput: AVCaptureVideoDataOutput = AVCaptureVideoDataOutput() var detectionRequests: [VNDetectFaceRectanglesRequest]? var videoDataOutputQueue: DispatchQueue = DispatchQueue(label: "queue") class XYZ: /*NSViewController or NSObject*/NSObject, AVCaptureVideoDataOutputSampleBufferDelegate { func viewDidLoad() { //super.viewDidLoad() let session = AVCaptureSession() let inputDevice = try! self.configureFrontCamera(for: session) self.configureVideoDataOutput(for: inputDevice.device, resolution: inputDevice.resolution, captureSession: session) self.prepareVisionRequest() session.startRunning() } fileprivate func highestResolution420Format(for device: AVCaptureDevice) -> (format: AVCaptureDevice.Format, resolution: CGSize)? { let deviceFormat = device.formats[0] print(deviceFormat) let dims = CMVideoFormatDescriptionGetDimensions(deviceFormat.formatDescription) let resolution = CGSize(width: CGFloat(dims.width), height: CGFloat(dims.height)) return (deviceFormat, resolution) } fileprivate func configureFrontCamera(for captureSession: AVCaptureSession) throws -> (device: AVCaptureDevice, resolution: CGSize) { let deviceDiscoverySession = AVCaptureDevice.DiscoverySession(deviceTypes: [AVCaptureDevice.DeviceType.builtInWideAngleCamera], mediaType: .video, position: AVCaptureDevice.Position.unspecified) let device = deviceDiscoverySession.devices.first! let deviceInput = try! AVCaptureDeviceInput(device: device) captureSession.addInput(deviceInput) let highestResolution = self.highestResolution420Format(for: device)! try! device.lockForConfiguration() device.activeFormat = highestResolution.format device.unlockForConfiguration() return (device, highestResolution.resolution) } fileprivate func configureVideoDataOutput(for inputDevice: AVCaptureDevice, resolution: CGSize, captureSession: AVCaptureSession) { videoDataOutput.setSampleBufferDelegate(self, queue: videoDataOutputQueue) captureSession.addOutput(videoDataOutput) } fileprivate func prepareVisionRequest() { let faceDetectionRequest: VNDetectFaceRectanglesRequest = VNDetectFaceRectanglesRequest(completionHandler: { (request, error) in print("VNDetectFaceRectanglesRequest completion Handler called") }) // Start with detection detectionRequests = [faceDetectionRequest] } // MARK: AVCaptureVideoDataOutputSampleBufferDelegate // Handle delegate method callback on receiving a sample buffer. public func captureOutput(_ output: AVCaptureOutput, didOutput sampleBuffer: CMSampleBuffer, from connection: AVCaptureConnection) { var requestHandlerOptions: [VNImageOption: AnyObject] = [:] let cameraIntrinsicData = CMGetAttachment(sampleBuffer, key: kCMSampleBufferAttachmentKey_CameraIntrinsicMatrix, attachmentModeOut: nil) if cameraIntrinsicData != nil { requestHandlerOptions[VNImageOption.cameraIntrinsics] = cameraIntrinsicData } let pixelBuffer = CMSampleBufferGetImageBuffer(sampleBuffer)! // No tracking object detected, so perform initial detection let imageRequestHandler = VNImageRequestHandler(cvPixelBuffer: pixelBuffer, orientation: CGImagePropertyOrientation.up, options: requestHandlerOptions) try! imageRequestHandler.perform(detectionRequests!) } } let X = XYZ() X.viewDidLoad() sleep(9999999)

I watched this year WWDC25 "Read Documents using the Vision framework". At the end of video there is mention of new DetectHandPoseRequest model for hand pose detection in Vision API. I looked Apple documentation and I don't see new revision. Moreover probably typo in video because there is only DetectHumanPoseRequst (swift based) and VNDetectHumanHandPoseRequest (obj-c based) (notice lack of Human prefix in WWDC video) First one have revision only added in iOS 18+: https://developer.apple.com/documentation/vision/detecthumanhandposerequest/revision-swift.enum/revision1 Second one have revision only added in iOS14+: https://developer.apple.com/documentation/vision/vndetecthumanhandposerequestrevision1 I don't see any new revision targeting iOS26+

After a few attempts I end up at the same spot as categorized here by ChatGPT - I get why it feels that way, Rob — honestly, you’ve walked right into the worst possible moment in Apple’s Core ML toolchain evolution. What you’re running into isn’t your fault; Apple really did rip out the ONNX front-end without offering a clean replacement yet. So every tutorial, every script, and even coremltools itself contradicts what’s actually possible right now. Let’s be brutally clear about the state of things (late 2025): Goal Supported by Apple today Practical workaround Convert ONNX → Core ML using coremltools 9 No. The ONNX importer was removed. Use the legacy coremltools 6.x + onnx-coreml stack in a separate Python 3.9/3.10 env. Convert PyTorch → Core ML directly Yes, but only for simple or traced models. T5 still fails because of dynamic control flow. None reliable without rewriting the model. Run ONNX models directly on iOS Partially (via MLC LLM / Metal Performance Shaders). Works, but that’s an entirely different runtime. So: you’re not crazy, and this isn’t wasted effort—you’ve actually proven that your model exports cleanly and that the only missing piece is Apple’s converter. The ONNX file you built is solid. You just need something that can translate it to Core ML. If your goal is simply “get it running on-device”, you still have two viable moves that aren’t rabbit holes: 1. Use MLC LLM / Metal runtime Skip Core ML altogether. Apple’s mlc-llm (and the “MLCChat” pipeline) can execute your ONNX or quantized model directly on the GPU. You keep full fidelity and avoid all these broken converters. 2. Use the legacy Core ML 6.x toolchain in a sandboxed Python 3.10 VM It actually works. You don’t have to downgrade your whole system—just run that specific environment once to emit the .mlpackage, then never touch it again. Nothing else will silently fix itself until Apple re-adds ONNX ingestion in Core ML 10 or a new converter.

The documentation for the Create ML tool ("Building an object detector data source") mentions that there are options for using normalized values instead of pixels and also different anchor point origins ("MLBoundingBoxCoordinatesOrigin") instead of always using "center". However, the JSON format for these does not appear in any examples. Does anyone know the format for these options?

Hi, I'm not sure whether this is the appropriate forum for this topic. I just followed a link from the JAX Metal plugin page https://developer.apple.com/metal/jax/ I'm writing a Python app with JAX, and recent JAX versions fail on Metal. E.g. v0.8.2 I have to downgrade JAX pretty hard to make it work: pip install jax==0.4.35 jaxlib==0.4.35 jax-metal==0.1.1 Can we get an updated release of jax-metal that would fix this issue? Here is the error I get with JAX v0.8.2: WARNING:2025-12-26 09:55:28,117:jax._src.xla_bridge:881: Platform 'METAL' is experimental and not all JAX functionality may be correctly supported! WARNING: All log messages before absl::InitializeLog() is called are written to STDERR W0000 00:00:1766771728.118004 207582 mps_client.cc:510] WARNING: JAX Apple GPU support is experimental and not all JAX functionality is correctly supported! Metal device set to: Apple M3 Max systemMemory: 36.00 GB maxCacheSize: 13.50 GB I0000 00:00:1766771728.129886 207582 service.cc:145] XLA service 0x600001fad300 initialized for platform METAL (this does not guarantee that XLA will be used). Devices: I0000 00:00:1766771728.129893 207582 service.cc:153] StreamExecutor device (0): Metal, <undefined> I0000 00:00:1766771728.130856 207582 mps_client.cc:406] Using Simple allocator. I0000 00:00:1766771728.130864 207582 mps_client.cc:384] XLA backend will use up to 28990554112 bytes on device 0 for SimpleAllocator. Traceback (most recent call last): File "<string>", line 1, in <module> import jax; print(jax.numpy.arange(10)) ~~~~~~~~~~~~~~~~^^^^ File "/Users/florin/git/FlorinAndrei/star-cluster-simulator/.venv/lib/python3.13/site-packages/jax/_src/numpy/lax_numpy.py", line 5951, in arange return _arange(start, stop=stop, step=step, dtype=dtype, out_sharding=sharding) File "/Users/florin/git/FlorinAndrei/star-cluster-simulator/.venv/lib/python3.13/site-packages/jax/_src/numpy/lax_numpy.py", line 6012, in _arange return lax.broadcasted_iota(dtype, (size,), 0, out_sharding=out_sharding) ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/florin/git/FlorinAndrei/star-cluster-simulator/.venv/lib/python3.13/site-packages/jax/_src/lax/lax.py", line 3415, in broadcasted_iota return iota_p.bind(dtype=dtype, shape=shape, ~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^ dimension=dimension, sharding=out_sharding) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/florin/git/FlorinAndrei/star-cluster-simulator/.venv/lib/python3.13/site-packages/jax/_src/core.py", line 633, in bind return self._true_bind(*args, **params) ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "/Users/florin/git/FlorinAndrei/star-cluster-simulator/.venv/lib/python3.13/site-packages/jax/_src/core.py", line 649, in _true_bind return self.bind_with_trace(prev_trace, args, params) ~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/Users/florin/git/FlorinAndrei/star-cluster-simulator/.venv/lib/python3.13/site-packages/jax/_src/core.py", line 661, in bind_with_trace return trace.process_primitive(self, args, params) ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^ File "/Users/florin/git/FlorinAndrei/star-cluster-simulator/.venv/lib/python3.13/site-packages/jax/_src/core.py", line 1210, in process_primitive return primitive.impl(*args, **params) ~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^ File "/Users/florin/git/FlorinAndrei/star-cluster-simulator/.venv/lib/python3.13/site-packages/jax/_src/dispatch.py", line 91, in apply_primitive outs = fun(*args) jax.errors.JaxRuntimeError: UNKNOWN: -:0:0: error: unknown attribute code: 22 -:0:0: note: in bytecode version 6 produced by: StableHLO_v1.13.0 -------------------- For simplicity, JAX has removed its internal frames from the traceback of the following exception. Set JAX_TRACEBACK_FILTERING=off to include these. I0000 00:00:1766771728.149951 207582 mps_client.h:209] MetalClient destroyed.

I’m trying to follow Apple’s “WWDC24: Bring your machine learning and AI models to Apple Silicon” session to convert the Mistral-7B-Instruct-v0.2 model into a Core ML package, but I’ve run into a roadblock that I can’t seem to overcome. I’ve uploaded my full conversion script here for reference: https://pastebin.com/T7Zchzfc When I run the script, it progresses through tracing and MIL conversion but then fails at the backend_mlprogram stage with this error: https://pastebin.com/fUdEzzKM The core of the error is: ValueError: Op "keyCache_tmp" (op_type: identity) Input x="keyCache" expects list, tensor, or scalar but got state[tensor[1,32,8,2048,128,fp16]] I’ve registered my KV-cache buffers in a StatefulMistralWrapper subclass of nn.Module, matching the keyCache and valueCache state names in my ct.StateType definitions, but Core ML’s backend pass reports the state tensor as an invalid input. I’m using Core ML Tools 8.3.0 on Python 3.9.6, targeting iOS18, and forcing CPU conversion (MPS wasn’t available). Any pointers on how to satisfy the handle_unused_inputs pass or properly declare/cache state for GQA models in Core ML would be greatly appreciated! Thanks in advance for your help, Usman Khan

Subject: Technical Report: Float32 Precision Ceiling & Memory Fragmentation in JAX/Metal Workloads on M3 To: Metal Developer Relations Hello, I am reporting a repeatable numerical saturation point encountered during sustained recursive high-order differential workloads on the Apple M3 (16 GB unified memory) using the JAX Metal backend. Workload Characteristics: Large-scale vector projections across multi-dimensional industrial datasets Repeated high-order finite-difference calculations Heavy use of jax.grad and lax.cond inside long-running loops Observation: Under these conditions, the Metal/MPS backend consistently enters a terminal quantization lock where outputs saturate at a fixed scalar value (2.0000), followed by system-wide NaN propagation. This appears to be a precision-limited boundary in the JAX-Metal bridge when handling high-order operations with cubic time-scale denominators. have identified the specific threshold where recursive high-order tensor derivatives exceed the numerical resolution of 32-bit consumer architectures, necessitating a migration to a dedicated 64-bit industrial stack. I have prepared a minimal synthetic test script (randomized vectors only, no proprietary logic) that reliably reproduces the allocator fragmentation and saturation behavior. Let me know if your team would like the telemetry for XLA/MPS optimization purposes. Best regards, Alex Severson Architect, QuantumPulse AI