Google AI Edge Gallery launches on macOS, letting Mac users run Gemini models locally
Google AI Edge Gallery is now on macOS, giving Mac users local access to selected Google models. Google also launched Gemma 4 12B and an on-device dictation app.
Intelligence analysis by GPT-5.4 Mini
Google is moving more AI work onto the Mac itself. The new macOS AI Edge Gallery supports five of Google’s own models, including Gemma 4 12B, and the companion dictation app processes speech locally instead of in the cloud.
Google put a small AI helper inside the Mac instead of making it live only on the internet. It is like having a notebook and a speech helper built into the computer, so some work can happen right there at home.
Analysis
What launched
Google released AI Edge Gallery for macOS, expanding a local-AI app it already offered on Android and iOS. On the Mac, the gallery does not act like broader local model managers such as Ollama or LM Studio; instead, it currently exposes only five Google models: Gemma-4-12B-it, Gemma-4-E2B-it, Gemma-4-E4B-it, Gemma-3n-E2B-it, and Gemma-3n-E4B-it.
Why Gemma 4 12B stands out
The article singles out Gemma 4 12B as the most notable launch. Google says it is designed to bring agentic, multimodal intelligence to laptops, and that its 12-billion-parameter design performs comparably to a 26-billion-parameter mixture-of-experts model while still being small enough to run locally on consumer laptops with 16GB of RAM. Google also says it can handle text, vision, and audio, and that it has coding capabilities that can help users extract insights from data on-device.
The second app: Eloquent
Google also launched AI Edge Eloquent for Mac, after bringing it to iOS earlier. Eloquent is a free dictation app that records speech, transcribes it, and then polishes the result by removing disfluencies and making light edits for clarity and flow. Because processing happens on-device, the app keeps transcription work off the cloud. It also includes writing-style options and custom word support, which should help reduce errors with names, jargon, and other repeated terms.
The broader shift
The piece frames local models as a practical alternative to cloud-based LLMs: they can work offline, run faster on better hardware, and keep conversation data on the device. Google’s macOS release fits that trend while keeping the experience tied to its own model lineup.
Key points
- Google AI Edge Gallery is now available on macOS after previously shipping on Android and iOS.
- The Mac version currently offers only five Google models, all of them instruction-tuned variants of Gemma.
- Google says Gemma 4 12B can run locally on consumer laptops with 16GB of RAM and handle text, vision, and audio.
- Google also launched AI Edge Eloquent for Mac, a local dictation app that transcribes and polishes speech on-device.
If the Mac version catches on, users could get faster responses and work without an internet connection. The dictation app could also appeal to people who want cleaner transcripts while keeping speech processing on the device.
The Mac app currently limits users to Google’s own models, unlike tools that let people choose from a wider range. That narrower menu could make it less attractive for users who want more flexibility or model variety.
