AnyLanguageModel 1.0: One Year of Local Models on Apple Platforms
huggingface ·
Hugging Face has released AnyLanguageModel 1.0, a Swift package that provides a drop-in compatible API for Apple platforms, allowing developers to easily switch between local and hosted inference providers. This update follows one year of development based on feedback from app developers seeking to simplify the integration of local models without managing multiple inference complexities. The new version supports tool calling, guided generation, and image attachment across all providers, including open-weight models via llama.cpp, MLX, and Core ML. Hosted model support has expanded significantly to include OpenAI, Anthropic, Google Gemini, Ollama, and Hugging Face Inference Providers, enabling the use of large models like GPT-OSS-120B through cloud endpoints. Local model capabilities have also improved with popular downloads such as Gemma 4 E2B and Qwen3.8 27B, which are optimized for various hardware constraints from iPhones to Mac Studios. The release coincides with swift-huggingface 1.0 and swift-xet 1.0, which accelerate model downloads from the Hub directly into Swift applications. Developers can now upgrade existing apps using migration guidance provided in the release notes, while future updates promise to extend this unified approach to decision-making and audio listening models.
huggingface