Hugging Face announced the 1.0 release of AnyLanguageModel, a Swift framework that provides a unified API for running language models across Apple platforms. The framework, which debuted last November, consolidates support for both local models running directly on devices and hosted services, eliminating complexity for app developers deciding between inference options. Over the past year, the project expanded to support multiple inference providers and added significant features, reaching production-ready status. The 1.0 release brings feature parity across all supported models and providers. Developers can now use tool calling and guided generation with any provider, attach images to prompts using Apple's Foundation Models syntax, and access models through OpenAI, Anthropic, Google Gemini, Ollama, Hugging Face Inference Providers, and OpenRouter with identical code. The framework also runs on Linux, enabling the same Swift codebase to work on servers and edge devices. As open-weight models have improved in capability and shrunk in size, small models like Gemma 4 from Google and Qwen3.8 from Alibaba now fit on iPhones or run efficiently on Macs. Hugging Face also released stable 1.0 versions of swift-huggingface and swift-xet libraries to accelerate model downloads, signaling confidence in the local-inference momentum on consumer devices.