Google expanded its AI Edge Gallery platform to macOS on June 3 2026, enabling users to run locally processed AI models on their Mac hardware. Unlike third-party platforms such as Ollama or LM Studio that support a wide variety of third-party models, the Google AI Edge Gallery for Mac is currently restricted to five of Google’s own models: Gemma-4-12B-it, Gemma-4-E2B-it, Gemma-4-E4B-it, Gemma-3n-E2B-it, and Gemma-3n-E4B-it. These models are designed to operate without an internet connection, utilizing the computer's onboard processing power to enhance privacy by ensuring data remains on the device.

A primary addition to the platform is the Gemma 4 12B model, which Google describes as capable of delivering performance comparable to a 26-billion-parameter mixture-of-experts (a model architecture that activates only specific parts of the network per input) model. Despite its 12-billion-parameter size, the model is optimized to run on consumer laptops equipped with 16GB of RAM. The model is also multimodal, providing the capability to process text, vision, and audio inputs while supporting coding tasks for local data analysis.

Additionally, Google released the Google AI Edge Eloquent app for macOS, a free dictation tool that performs transcription locally on the device. The application transcribes spoken input while automatically refining the text to remove disfluencies and improve clarity. Users can customize the app to recognize specific jargon or names, reducing common dictation errors. This release follows the app’s debut on iOS several months ago.

Google expanded its AI Edge Gallery platform to macOS on June 3 2026, enabling users to run locally processed AI models on their Mac hardware. Unlike third-party platforms such as Ollama or LM Studio that support a wide variety of third-party models, the Google AI Edge Gallery for Mac is currently restricted to five of Google’s own models: Gemma-4-12B-it, Gemma-4-E2B-it, Gemma-4-E4B-it, Gemma-3n-E2B-it, and Gemma-3n-E4B-it. These models are designed to operate without an internet connection, utilizing the computer's onboard processing power to enhance privacy by ensuring data remains on the device.

A primary addition to the platform is the Gemma 4 12B model, which Google describes as capable of delivering performance comparable to a 26-billion-parameter mixture-of-experts (a model architecture that activates only specific parts of the network per input) model. Despite its 12-billion-parameter size, the model is optimized to run on consumer laptops equipped with 16GB of RAM. The model is also multimodal, providing the capability to process text, vision, and audio inputs while supporting coding tasks for local data analysis.

Additionally, Google released the Google AI Edge Eloquent app for macOS, a free dictation tool that performs transcription locally on the device. The application transcribes spoken input while automatically refining the text to remove disfluencies and improve clarity. Users can customize the app to recognize specific jargon or names, reducing common dictation errors. This release follows the app’s debut on iOS several months ago.