Google Reveals RAM Requirements for Running AI Locally on Phones
Google has shed light on the significant RAM requirements needed to run its Gemini AI model locally on mobile devices, particularly on its Pixel 8 smartphones.
During a recent episode of the Made by Google podcast, Seang Chau, VP of devices and services software, disclosed that while the Pixel 8 Pro’s ample 12 GB of RAM made it suitable for hosting Gemini Nano, the standard Pixel 8’s 8 GB fell short of Google’s standards for an optimal user experience.
However, Google has now announced plans to introduce Gemini Nano to Pixel 8 users as a developer option in an upcoming software update. This move aims to provide developers with access to the AI model while acknowledging the hardware limitations of the Pixel 8.
Chau emphasized Google’s desire for AI-enabled features like smart reply to be “RAM resident,” ensuring their readiness for immediate use. While this approach offers convenience, it raises concerns about the permanent allocation of system memory and its potential impact on device performance.
Notably, Google clarified to Ars Technica that neither the Pixel 8 nor the Pixel 8 Pro currently retain Gemini in memory by default, requiring users to enable a developer flag to do so.
As users weigh the benefits of on-device AI functionality against potential performance implications, Google’s efforts to optimize AI experiences on mobile devices continue to evolve, reflecting the delicate balance between innovation and practicality in smartphone technology.
