Google has launched its AI Edge Gallery app on Mac for the first time, allowing users to run Gemini large language models locally without internet. Previously available only on iPhone, this release brings advanced AI capabilities directly to Mac users. Running LLMs locally offers privacy benefits and faster performance than cloud-based alternatives, addressing growing concerns about data privacy and AI responsiveness in consumer applications.

Alongside the app, Google released the Gemma 4 12B model designed for local execution on laptops. The model requires at least 16GB of VRAM or unified memory, making it compatible with all modern Apple laptops except the MacBook Neo. This "agentic multimodal intelligence" approach represents Google's strategy to make more powerful AI models accessible on personal devices without cloud dependency, potentially reducing latency and improving user experience.

Google also introduced AI Edge Eloquent, an AI-powered dictation and editing tool that works across all Mac apps without internet connection. The free tool, previously available on iPhone, allows users to choose writing styles and create custom vocabularies. While currently English-only, more languages are planned. This launch signals Google's commitment to expanding its on-device AI presence on Apple's platform, addressing the growing demand for private, offline AI solutions.