Edgar Cervantes from Android Authority reports that Google has recently added support for Gemini Nano on the latest Chrome Canary builds. This feature is hidden behind two Chrome flags and requires the manual download of the language model. Gemini Nano operates offline with response times of less than one second for simple queries on modern computers. Although the model is small enough to run on a Pixel 8, its use cases are limited to features like Gboard’s Smart Reply or AI-generated summaries in the Recorder app.
Google has now begun testing Gemini Nano integration within Chrome for desktops. This integration allows users to interact with a large language model entirely within the web browser, even offline. The announcement for Gemini Nano for Chrome was made last month, with plans to make it accessible to developers for testing. It was spotted in version 127 of Chrome Canary a few weeks ago, and developers have already created web apps showcasing the model’s capabilities.
To enable Gemini Nano in Chrome, users must follow a few steps such as enabling specific feature flags and waiting for the model to download. Although Google has not provided an interface to chat with Gemini Nano, an open-source web app created by Lightning Joyce can be used. While testing the model on a mid-range PC, the performance was impressive, with the model responding quickly to prompts.
However, there were instances where Gemini Nano failed to respond to certain open-ended prompts, causing a spike in GPU usage. Despite these issues, Gemini Nano remains quick and usable, only taking up 2GB of GPU video memory. The model’s performance within Chrome makes it easily accessible, with steps provided for users to try it on their own machines. Those interested in using Gemini Nano on Chrome can follow the steps outlined above.
For any tips or information, readers can contact the Android Authority staff at news@androidauthority.com. They have the option to remain anonymous or receive credit for their input.