ChromeAI

This page is a practical test environment for Google Chrome’s built‑in large language model, Gemini Nano, which runs entirely inside the user’s browser on their local machine. No data is sent to any external server; all computation happens client‑side, ensuring user privacy and offline capability once the model is loaded. The interface presents a straightforward chatbot‑style interaction where users type a question or prompt, and Gemini Nano generates a response directly in the browser. Additionally, a **summarization** feature is available: users can paste a block of text, and the model will produce a concise summary of its content. The system also includes a **model comparison** tool that allows users to send the same prompt to ChromeAI (Gemini Nano), ChatGPT, Claude, and Llama (via external APIs where those models are hosted) and view the results side by side. This is particularly useful for developers, researchers, or AI enthusiasts who want to evaluate the performance, style, and accuracy of a locally‑run small language model against popular cloud‑based alternatives. To use the page, one must have Google Chrome version 127 or higher, as older versions do not include the required Gemini Nano support. Additionally, several experimental browser flags need to be manually enabled through `chrome://flags` — the page provides a quick setup guide that lists the exact flags (e.g., “Prompt API for Gemini Nano” and “Enables the Summarization API”) and explains how to turn them on. Once these flags are activated and Chrome is relaunched, the model downloads automatically in the background (typically a few hundred megabytes of data) and becomes ready for local inference. The entire user history, including past conversations and summaries, is stored locally in the browser’s Inde