New Tool Lets iPhone Apps Run Powerful AI Without Internet
Your favorite apps could soon work smarter offline, saving data and protecting privacy.
AnyLanguageModel 1.0 is out, a year and dozens of releases after it was introduced on the blog as a drop-in replacement for Apple's Foundation Models framework. It grew out of feedback from app developers who were curious about local models but weren't sure how to get started, and who wanted to avoid the complexity of supporting multiple inference providers.
The core idea is unchanged: use the Foundation Models API with the model that fits your app, starting by changing your import statement. The same session code then works with the system model, with open-weight models through llama.cpp, MLX, and Core ML, and with hosted providers.
Alongside it, swift-huggingface 1.0 and swift-xet 1.0 bring accelerated model downloads from the Hub to Swift apps. Hosted providers get the same features as local models: guided generation, tool calling with streaming, and token usage on responses.
- AnyLanguageModel 1.0 lets developers use AI that runs on your device, so apps work offline and keep your data private.
- Some AI models are now small enough to fit on an iPhone, like Qwen3.8, making local AI practical for everyday apps.
- The tool also supports cloud-based AI for heavier tasks, giving developers flexibility to choose the best option.
Why It Matters
Apps will become smarter, faster, and more private, working offline and saving data.