Chrome Prompt API Integrates On-Device Gemini Nano
Chrome Prompt API enables on-device natural language prompting with Gemini Nano under defined hardware and OS requirements.
Chrome has released the Prompt API to enable natural language requests to the Gemini Nano model directly in the browser.
The API requires Windows 10 or 11, macOS 13+, Linux or ChromeOS on Chromebook Plus devices with at least 22 GB free storage, GPU exceeding 4 GB VRAM or CPU with 16 GB RAM and 4 cores; audio input requires GPU (https://developer.chrome.com/docs/ai/prompt-api). Model downloads on first origin use after LanguageModel.availability() check, with downloadprogress events supported.
Supported applications listed include AI-powered search on page content, dynamic news feed classification, custom content filters, calendar event extraction and contact information processing (https://developer.chrome.com/docs/ai/prompt-api). Parameters reported by LanguageModel.params() list defaultTopK 3, maxTopK 128, defaultTemperature 1 and maxTemperature 2.
Google introduced Gemini Nano in December 2023 for on-device Android features including summarization (https://blog.google/technology/ai/gemini-nano-android/). The Prompt API aligns with W3C Web Neural Network API efforts to standardize browser ML inference without cloud dependencies (https://www.w3.org/TR/webnn/).
AXIOM: Chrome's Prompt API accelerates on-device generative AI adoption by enabling local prompting with Gemini Nano, improving privacy through eliminated cloud calls and supporting new browser applications independent of remote servers.
Sources (3)
- [1]The Prompt API(https://developer.chrome.com/docs/ai/prompt-api)
- [2]Gemini Nano for Android(https://blog.google/technology/ai/gemini-nano-android/)
- [3]Web Neural Network API(https://www.w3.org/TR/webnn/)