According to Observant Beating monitoring, LM Studio has officially launched its official mobile app Locally (available on iPhone and iPad). Earlier this year, LM Studio completed the acquisition of the Locally app, and this update deeply integrates it into its own ecosystem, providing users with the ability to use local AI large models on mobile devices. The highlight of Locally lies in its support for both "on-device local execution" and remote desktop computing through the new "LM Link" feature, breaking the physical boundaries of personal device computing power.
Locally has been deeply optimized for Apple Silicon, built on Apple's official MLX machine learning framework to achieve high performance and energy efficiency, and natively supports Apple Foundation models. The app currently supports mainstream open-source models such as Llama 3.2, Google Gemma 4 series, Qwen 3, and DeepSeek, allowing users to directly run models like Gemma 4 (E2B) on their phones and freely import custom GGUF format models. Locally adheres to a 100% offline and privacy protection principle, with all on-device chat and processing processes completed locally.
To enable users to use large-parameter cutting-edge models on mobile devices, Locally introduces the LM Link remote connection feature. Enabling this feature requires users to log in to the same LM Studio account on both their phone and computer for device authorization. After completing the login and establishing an exclusive network connection, this feature will build an encrypted network based on Tailscale's mesh VPN technology. Leveraging the virtual private encrypted channel established by Tailscale, users can securely connect to their home or office desktop computers from an iPhone at any time and remote call large-parameter models such as Gemma 4 (31B) running on desktop devices like the MacBook Pro (M5 Max). This architecture ensures that all data transfers between devices are end-to-end encrypted, with no device exposed to the public network, balancing mobile convenience with ultimate local privacy.
The integration of Locally and LM Link has opened up a new workflow for local AI developers and enthusiasts. Users can transform their desktop hardware assets into their own "portable AI brain" without incurring expensive cloud API fees or facing privacy risks.
