According to Beating, LM Studio officially launched Locally, its mobile application for iPhone and iPad. The app combines local on-device AI model execution with LM Link, a new remote access feature that allows users to connect their iPhones to desktop computers running larger models. The connection is secured through Tailscale's mesh VPN technology, enabling encrypted end-to-end communication without exposing devices to the public internet.
Locally supports major open-source models including Llama 3.2, Google Gemma 4, Qwen 3, and DeepSeek. The app is optimized for Apple Silicon chips using Apple's MLX machine learning framework. Users can run models locally on-device while offline or use LM Link to access desktop models like Gemma 4 (31B) running on MacBook Pro (M5 Max) from anywhere.