Google's experimental app AI Edge Gallery has officially launched on Apple macOS. This move means that Mac users can now run the latest Gemma series of artificial intelligence models locally on their laptops without relying on a network connection.

Thanks to its fully offline operation, the app not only significantly improves response speed but also ensures absolute data privacy for users. Users can smoothly conduct intelligent conversations, image processing, and deep semantic understanding without being online.

image.png

First Release of Multimodal Models

In the five official models now available for Mac, the most notable is Gemma-4-12B-it. Google emphasized that this model has been deeply optimized and can run smoothly on Mac devices with just 16GB of memory.

It can easily handle information from multiple modalities such as text, vision, and audio, and also has a strong ability to write and understand code. This allows ordinary laptops to become powerful multimodal intelligences, performing local data analysis anytime, anywhere.

Offline Efficient Speech Recognition Launched at the Same Time

Alongside this gallery, there is also a free voice transcription app called Google AI Edge Eloquent. It is specifically used to capture user speech and convert it into text in real time, and it can automatically filter out mistakes and perform light polishing on the text.

To prevent industry terms or names from being incorrectly modified by the software, it also supports custom vocabulary libraries and different writing styles. It is worth noting that all audio and text processing in this software is done locally on the device, and nothing is uploaded to the cloud.