Particle.news
Download on the App Store
3 ARTICLES
·
2w ago

Google Brings Gemma Models and On‑Device AI Tools to macOS

The macOS launch lets Macs run a new multimodal 12‑billion‑parameter model and a local dictation app so data stays on the device, enables offline use, and cuts response lag.

Use Google's new Mac app to run its AI models locally.
Run Google's Gemini LLMs right on your Mac with AI Edge Gallery
Image

Overview

  • Google released the AI Edge Gallery app for macOS on Thursday, June 4, 2026, and published the Gemma‑4‑12B model and the AI Edge Eloquent dictation app for Mac.
  • Gemma‑4‑12B is a unified, encoder‑free multimodal model designed to run on consumer laptops with about 16GB of RAM while handling text, audio, and vision inputs.
  • The macOS AI Edge Gallery runs processing on Apple Silicon and exposes five of Google’s own Gemma ‘it’ instruct models for local use rather than letting users install third‑party models.
  • AI Edge Eloquent is a free on‑device dictation and editing app that transcribes, removes disfluencies, and polishes text inside any Mac app via a shortcut key with no data sent to Google’s servers.
  • The move follows a broader shift toward local LLMs that trade peak cloud scale for privacy, offline capability, and lower latency and could let technically skilled users build multi‑turn local agents while cloud assistants remain widely used.