AnyLanguageModel 1.0: one year of local models on Apple platforms
Hugging Face releases AnyLanguageModel 1.0, a drop-in replacement for Apple's Foundation Models framework that exposes local or hosted models through one API, alongside swift-huggingface 1.0 and swift-xet 1.0. It adds tool calling, guided generation, image support, many hosted providers and Linux support.
What the article says
AnyLanguageModel is a drop-in replacement for Apple's Foundation Models framework that lets developers choose which model and provider their app uses. The project began in November 2025 from developer feedback and reaches version 1.0 alongside swift-huggingface 1.0 and swift-xet 1.0, two packages that speed up model downloads from the Hugging Face Hub in Swift apps. Migrating requires only changing the FoundationModels import to AnyLanguageModel.
The same session code works with the system model, open-weight models through llama.cpp, MLX and Core ML, and hosted providers such as OpenAI, Anthropic, Google Gemini or Ollama. Tool calling and guided generation work with every provider, and the library also runs on Linux. The post cites examples such as Gemma 4 E2B (about 3.4 GB) and LFM2.5 2.6B (about 1.7 GB), which fit on an iPhone, and Qwen3.8 27B (about 16 GB) for a Mac with 32 GB or more. The authors say other model types are coming soon, with no dates given.
Read the official article AnnuIA profile
Summary written by AnnuIA in its own words from the official page of Hugging Face. The content and rights belong to the publisher: please read the original article.