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.

Hello It seems the model Content Tagging doesn't obey when I define the type of tag I wish in the instructions parameters, always the output are the main topics. The unique form to get other type of tags like emotions is using Generable + Guided types. The documentation says it is recommended but not mandatory the use instructions. Maybe I'm setting wrongly the instructions but take a look in the attached snapshot. I copied the definition of tagging emotions from the official documentation. The upper example is employing generable and it works but in the example at the botton I set like instruction the same description of emotion and it doesn't work. I tried with other statements with more or less verbose and never output emotions. Could you provide a state using instruction where it works? Current version of model isn't working with instruction?

I am writing to inquire about content exclusion capabilities within Apple Intelligence, particularly regarding the use of configuration files such as .aiignore or .aiexclude—similar to what exists in other AI-assisted coding tools. These mechanisms are highly valuable in managing what content AI systems can access, especially in environments that involve sensitive code or proprietary frameworks. I would appreciate it if anyone could clarify whether Apple Intelligence currently supports any exclusion configuration for AI-assisted features. If so, could you kindly provide documentation or guidance on how developers can implement these controls? If not, Is there any plan to include such feature in future updates?

Hi everyone, I'm developing an iOS app using Foundation Models and I've hit a critical limitation that I believe affects many developers and millions of users. The Issue Foundation Models requires the device system language to be one of the supported languages. If a user has their device set to an unsupported language (Catalan, Dutch, Swedish, Polish, Danish, Norwegian, Finnish, Czech, Hungarian, Greek, Romanian, and many others), SystemLanguageModel.isSupported returns false and the framework is completely unavailable. Why This Is Problematic Scenario: A Catalan user has their iPhone in Catalan (native language). They want to use an AI chat app in Spanish or English (languages they speak fluently). Current situation: ❌ Foundation Models: Completely unavailable ✅ OpenAI GPT-4: Works perfectly ✅ Anthropic Claude: Works perfectly ✅ Any cloud-based AI: Works perfectly The user must choose between: Keep device in Catalan → Cannot use Foundation Models at all Change entire device to Spanish → Can use Foundation Models but terrible UX Impact This affects: Millions of users in regions where unsupported languages are official Multilingual users who prefer their device in their native language but can comfortably interact with AI in English/Spanish Developers who cannot deploy Foundation Models-based apps in these markets Privacy-conscious users who are ironically forced to use cloud AI instead of on-device AI What We Need One of these solutions would solve the problem: Option 1: Per-app language override (preferred) // Proposed API let session = try await LanguageModelSession(preferredLanguage: "es-ES") Option 2: Faster rollout of additional languages (particularly EU languages) Option 3: Allow fallback to user-selected supported language when system language is unsupported Technical Details Current behavior: // Device in Catalan let isAvailable = SystemLanguageModel.isSupported // Returns false // No way to override or specify alternative language Why This Matters Apple Intelligence and Foundation Models are amazing for privacy and performance. But this language restriction makes the most privacy-focused AI solution less accessible than cloud alternatives. This seems contrary to Apple's values of accessibility and user choice. Questions for the Community Has anyone else encountered this limitation? Are there any workarounds I'm missing? Has anyone successfully filed feedback about this?(Please share FB number so we can reference it) Are there any sessions or labs where this has been discussed? Thanks for reading. I'd love to hear if others are facing this and how you're handling it.

Lately I am getting this error. GenerativeModelsAvailability.Parameters: Initialized with invalid language code: en-GB. Expected to receive two-letter ISO 639 code. e.g. 'zh' or 'en'. Falling back to: en Does anyone know what this is and how it can be resolved. The error does not crash the app

Using Tensorflow for Silicon gives inaccurate results when compared to Google Colab GPU (9-15% differences). Here are my install versions for 4 anaconda env's. I understand the Floating point precision can be an issue, batch size, activation functions but how do you rectify this issue for the past 3 years? 1.) Version TF: 2.12.0, Python 3.10.13, tensorflow-deps: 2.9.0, tensorflow-metal: 1.2.0, h5py: 3.6.0, keras: 2.12.0 2.) Version TF: 2.19.0, Python 3.11.0, tensorflow-metal: 1.2.0, h5py: 3.13.0, keras: 3.9.2, jax: 0.6.0, jax-metal: 0.1.1,jaxlib: 0.6.0, ml_dtypes: 0.5.1 3.) python: 3.10.13,tensorflow: 2.19.0,tensorflow-metal: 1.2.0, h5py: 3.13.0, keras: 3.9.2, ml_dtypes: 0.5.1 4.) Version TF: 2.16.2, tensorflow-deps:2.9.0,Python: 3.10.16, tensorflow-macos 2.16.2, tensorflow-metal: 1.2.0, h5py:3.13.0, keras: 3.9.2, ml_dtypes: 0.3.2 Install of Each ENV with common example: Create ENV: conda create --name TF_Env_V2 --no-default-packages start env: source TF_Env_Name ENV_1.) conda install -c apple tensorflow-deps , conda install tensorflow,pip install tensorflow-metal,conda install ipykernel ENV_2.) conda install pip python==3.11, pip install tensorflow,pip install tensorflow-metal,conda install ipykernel ENV_3) conda install pip python 3.10.13,pip install tensorflow, pip install tensorflow-metal,conda install ipykernel ENV_4) conda install -c apple tensorflow-deps, pip install tensorflow-macos, pip install tensor-metal, conda install ipykernel Example used on all 4 env: import tensorflow as tf cifar = tf.keras.datasets.cifar100 (x_train, y_train), (x_test, y_test) = cifar.load_data() model = tf.keras.applications.ResNet50( include_top=True, weights=None, input_shape=(32, 32, 3), classes=100,) loss_fn = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False) model.compile(optimizer="adam", loss=loss_fn, metrics=["accuracy"]) model.fit(x_train, y_train, epochs=5, batch_size=64)

Context I’m deploying large language models on iPhone using llama.cpp. A new iPhone Air (12 GB RAM) reports a Metal MTLDevice.recommendedMaxWorkingSetSize of 8,192 MB, and my attempt to load Llama-2-13B Q4_K (~7.32 GB weights) fails during model initialization. Environment Device: iPhone Air (12 GB RAM) iOS: 26 Xcode: 26.0.1 Build: Metal backend enabled llama.cpp App runs on device (not Simulator) What I’m seeing MTLCreateSystemDefaultDevice().recommendedMaxWorkingSetSize == 8192 MiB Loading Llama-2-13B Q4_K (7.32 GB) fails to complete. Logs indicate memory pressure / allocation issues consistent with the 8 GB working-set guidance. Smaller models (e.g., 7B/8B with similar quantization) load and run (8B Q4_K provide around 9 tokens/second decoding speed). Questions Is 8,192 MB an expected recommendedMaxWorkingSetSize on a 12 GB iPhone? What values should I expect on other 2025 devices including iPhone 17 (8 GB RAM) and iPhone 17 Pro (12 GB RAM) Is it strictly enforced by Metal allocations (heaps/buffers), or advisory for best performance/eviction behavior? Can a process practically exceed this for long-lived buffers without immediate Jetsam risk? Any guidance for LLM scenarios near the limit?

I'm experimenting with using the Foundation Models framework to do news summarization in an RSS app but I'm finding that a lot of articles are getting kicked back with a vague message about guardrails. This seems really common with political news but we're talking mainstream stuff, i.e. Politico, etc. If the models are this restrictive, this will be tough to use. Is this intended? FB17904424

Also submitted as feedback (ID: FB20612561). Tensorflow-metal fails on tensorflow versions above 2.18.1, but works fine on tensorflow 2.18.1 In a new python 3.12 virtual environment: pip install tensorflow pip install tensor flow-metal python -c "import tensorflow as tf" Prints error: Traceback (most recent call last): File "", line 1, in File "/Users//pt/venv/lib/python3.12/site-packages/tensorflow/init.py", line 438, in _ll.load_library(_plugin_dir) File "/Users//pt/venv/lib/python3.12/site-packages/tensorflow/python/framework/load_library.py", line 151, in load_library py_tf.TF_LoadLibrary(lib) tensorflow.python.framework.errors_impl.NotFoundError: dlopen(/Users//pt/venv/lib/python3.12/site-packages/tensorflow-plugins/libmetal_plugin.dylib, 0x0006): Library not loaded: @rpath/_pywrap_tensorflow_internal.so Referenced from: <8B62586B-B082-3113-93AB-FD766A9960AE> /Users//pt/venv/lib/python3.12/site-packages/tensorflow-plugins/libmetal_plugin.dylib Reason: tried: '/Users//pt/venv/lib/python3.12/site-packages/tensorflow-plugins/../_solib_darwin_arm64/_U@local_Uconfig_Utf_S_S_C_Upywrap_Utensorflow_Uinternal___Uexternal_Slocal_Uconfig_Utf/_pywrap_tensorflow_internal.so' (no such file), '/Users//pt/venv/lib/python3.12/site-packages/tensorflow-plugins/../_solib_darwin_arm64/_U@local_Uconfig_Utf_S_S_C_Upywrap_Utensorflow_Uinternal___Uexternal_Slocal_Uconfig_Utf/_pywrap_tensorflow_internal.so' (no such file), '/opt/homebrew/lib/_pywrap_tensorflow_internal.so' (no such file), '/System/Volumes/Preboot/Cryptexes/OS/opt/homebrew/lib/_pywrap_tensorflow_internal.so' (no such file)

Hello, is it allowed to use Foundation Model Framework in submission app for WWDC26? The thing is that Apple Intelligence needs to be enabled in the settings. So, does that mean the jury won't be able to fully utilize the app's AI functionality?

I am using macOS Tahoe on Xcode 26.

We’ve encountered what appears to be a CoreML regression between macOS 26.0.1 and macOS 26.1 Beta. In macOS 26.0.1, CoreML models run and produce correct results. However, in macOS 26.1 Beta, the same models produce scrambled or corrupted outputs, suggesting that tensor memory is being read or written incorrectly. The behavior is consistent with a low-level stride or pointer arithmetic issue — for example, using 16-bit strides on 32-bit data or other mismatches in tensor layout handling. Reproduction Install ON1 Photo RAW 2026 or ON1 Resize 2026 on macOS 26.0.1. Use the newest Highest Quality resize model, which is Stable Diffusion–based and runs through CoreML. Observe correct, high-quality results. Upgrade to macOS 26.1 Beta and run the same operation again. The output becomes visually scrambled or corrupted. We are also seeing similar issues with another Stable Diffusion UNet model that previously worked correctly on macOS 26.0.1. This suggests the regression may affect multiple diffusion-style architectures, likely due to a change in CoreML’s tensor stride, layout computation, or memory alignment between these versions. Notes The affected models are exported using standard CoreML conversion pipelines. No custom operators or third-party CoreML runtime layers are used. The issue reproduces consistently across multiple machines. It would be helpful to know if there were changes to CoreML’s tensor layout, precision handling, or MLCompute backend between macOS 26.0.1 and 26.1 Beta, or if this is a known regression in the current beta.

Hello folks! Taking a look at https://developer.apple.com/documentation/foundationmodels it’s not clear how to use another models there. Do anyone knows if it’s possible use one trained model from outside (imported) here in foundation models framework? Thanks!

Hello! I'm following the Foundation Models adapter training guide (https://developer.apple.com/apple-intelligence/foundation-models-adapter/) on my NVIDIA DGX Spark box. I'm able to train on my own data but the example notebook fails when I try to export the artifact as an fmadapter. I get the following error for the code block I'm trying to run. I haven't touched any of the code in the export folder. I tried exporting it on my Mac too and got the same error as well (given below). Would appreciate some more clarity around this. Thank you. Code Block: from export.export_fmadapter import Metadata, export_fmadapter metadata = Metadata( author="3P developer", description="An adapter that writes play scripts.", ) export_fmadapter( output_dir="./", adapter_name="myPlaywritingAdapter", metadata=metadata, checkpoint="adapter-final.pt", draft_checkpoint="draft-model-final.pt", ) Error: --------------------------------------------------------------------------- ModuleNotFoundError Traceback (most recent call last) Cell In[10], line 1 ----> 1 from export.export_fmadapter import Metadata, export_fmadapter 3 metadata = Metadata( 4 author="3P developer", 5 description="An adapter that writes play scripts.", 6 ) 8 export_fmadapter( 9 output_dir="./", 10 adapter_name="myPlaywritingAdapter", (...) 13 draft_checkpoint="draft-model-final.pt", 14 ) File /workspace/export/export_fmadapter.py:11 8 from typing import Any 10 from .constants import BASE_SIGNATURE, MIL_PATH ---> 11 from .export_utils import AdapterConverter, AdapterSpec, DraftModelConverter, camelize 13 logger = logging.getLogger(__name__) 16 class MetadataKeys(enum.StrEnum): File /workspace/export/export_utils.py:15 13 import torch 14 import yaml ---> 15 from coremltools.libmilstoragepython import _BlobStorageWriter as BlobWriter 16 from coremltools.models.neural_network.quantization_utils import _get_kmeans_lookup_table_and_weight 17 from coremltools.optimize._utils import LutParams ModuleNotFoundError: No module named 'coremltools.libmilstoragepython'

Hello, I have created this basic swift program: let session = LanguageModelSession( model: .default, instructions: "bla bla bla.") I want to understand what I can put in model parameter (instead of .default). How can I choose between on-device local model (.default I suppose) and apple private cloud model (or any other ?) Thanks

I'm really not familiar with ML, but I need a model that can enhance and denoise 4k video stream at 30fps. I have tried to search latest papers but they all have very complex structure, and I don't think I can convert them to mlmodel. So can anyone give me any recommandation for such models? If there is an existing mlmodel, that would be great!

Hi all, I'm tuning my app prediction speed with Core ML model. I watched and tried the methods in video: Improve Core ML integration with async prediction and Optimize your Core ML usage. I also use instruments to look what's the bottleneck that my prediction speed cannot be faster. Below is the instruments result with my app. its prediction duration is 10.29ms And below is performance report shows the average speed of prediction is 5.55ms, that is about half time of my app prediction! Below is part of my instruments records. I think the prediction should be considered quite frequent. Could it be faster? How to be the same prediction speed as performance report? The prediction speed on macbook Pro M2 is nearly the same as macbook Air M1!

I have an app that streams in data from the Foundation Model and I have a card that shows one of the outputs. I want my card to accept a partially generated model but I keep getting a nonsensical error. The error I get on line 59 is: Cannot convert value of type 'FrostDate.VegetableSuggestion.PartiallyGenerated' (aka 'FrostDate.VegetableSuggestion') to expected argument type 'FrostDate.VegetableSuggestion.PartiallyGenerated' Here is my card with preview: import SwiftUI import FoundationModels struct VegetableSuggestionCard: View { let vegetableSuggestion: VegetableSuggestion.PartiallyGenerated init(vegetableSuggestion: VegetableSuggestion.PartiallyGenerated) { self.vegetableSuggestion = vegetableSuggestion } var body: some View { VStack(alignment: .leading, spacing: 8) { if let name = vegetableSuggestion.vegetableName { Text(name) .font(.headline) .frame(maxWidth: .infinity, alignment: .leading) } if let startIndoors = vegetableSuggestion.startSeedsIndoors { Text("Start indoors: \(startIndoors)") .frame(maxWidth: .infinity, alignment: .leading) } if let startOutdoors = vegetableSuggestion.startSeedsOutdoors { Text("Start outdoors: \(startOutdoors)") .frame(maxWidth: .infinity, alignment: .leading) } if let transplant = vegetableSuggestion.transplantSeedlingsOutdoors { Text("Transplant: \(transplant)") .frame(maxWidth: .infinity, alignment: .leading) } if let tips = vegetableSuggestion.tips { Text("Tips: \(tips)") .foregroundStyle(.secondary) .frame(maxWidth: .infinity, alignment: .leading) } } .padding(16) .frame(maxWidth: .infinity, alignment: .leading) .background( RoundedRectangle(cornerRadius: 16, style: .continuous) .fill(.background) .overlay( RoundedRectangle(cornerRadius: 16, style: .continuous) .strokeBorder(.quaternary, lineWidth: 1) ) .shadow(color: Color.black.opacity(0.05), radius: 6, x: 0, y: 2) ) } } #Preview("Vegetable Suggestion Card") { let sample = VegetableSuggestion.PartiallyGenerated( vegetableName: "Tomato", startSeedsIndoors: "6–8 weeks before last frost", startSeedsOutdoors: "After last frost when soil is warm", transplantSeedlingsOutdoors: "1–2 weeks after last frost", tips: "Harden off seedlings; provide full sun and consistent moisture." ) VegetableSuggestionCard(vegetableSuggestion: sample) .padding() .previewLayout(.sizeThatFits) }

The WWDC25: Explore large language models on Apple silicon with MLX video talks about using your own data to fine-tune a large language model. But the video doesn't explain what kind of data can be used. The video just shows the command to use and how to point to the data folder. Can I use PDFs, Word documents, Markdown files to train the model? Are there any code examples on GitHub that demonstrate how to do this?

Hi For certain tasks, such as qualitative analysis or tagging, it is advisable to provide the AI with the option to respond with a joker / wild card answer when it encounters difficulties in tagging or scoring. For instance, you can include this slot in the prompt as follows: output must be "not data to score" when there isn't information to score. In the absence of these types of slots, AI trends to provide a solution even when there is insufficient information. Foundations Models are told to be prompted with simple prompts. I wonder: Is recommended keep this slot though adds verbose complexity? Is the best place the comment of a guided attribute? other tips? Another use case is when you want the AI to be tied to the information provided in the prompt and not take information from its data set. What is the best approach to this purpose? Thanks in advance for any suggestion.

Hey dear developers! This post should be available for the future Siri updates and improvements but also for wishes in this forum so that everyone can share their opinion and idea please stay friendly. have fun! I had already thought about developing a demo app to demonstrate my idea for a better Siri. My change of many: Wish Update: Siri's language recognition capabilities have been significantly enhanced. Instead of manually setting the language, Siri can now automatically recognize the language you intend to use, making language switching much more efficient. Simply speak the language you want to communicate in, and Siri will automatically recognize it and respond accordingly. Whether you speak English, German, or Japanese, Siri will respond in the language you choose.