**Gemini Nano to be Pre-installed on Chrome**

A quantized Gemini model with 325 million parameters will soon run locally on Google Chrome, with a latency of under 100 milliseconds and memory usage of less than 2GB.

This shows that it uses less memory than many current Chrome pages.

In the future, LLMs are expected to be usable not only in browsers but also in TVs, home appliances, and anywhere else.

The most interesting aspect is that all these models will run locally, making them fast and decentralized.