Chrome’s Prompt API brings on-device LLM inference to the browser, letting web pages run prompts locally without external API calls.
Key Takeaways
Part of Chrome’s “AI on Chrome” initiative, exposing a built-in model to JavaScript running in the browser.
Prompts run against a local model on the user’s device, keeping data off external servers by default.
No backend infrastructure or API keys required from the developer side; the browser handles model execution.
Targets a broad surface: any web page or extension can call the Prompt API if the user has Chrome with the model available.
Hacker News Comment Review
The dominant concern is on-device model size: users face a ~22 GB disk space requirement before any AI-on-Chrome feature works, which is a hard gate for many machines.
This storage barrier shifts the friction from API keys and latency to local hardware constraints, making the “no backend needed” pitch conditional on users having modern, high-capacity devices.
Notable Comments
@fg137: highlights the user-hostile implication of the storage gate: “sorry, to use our website, you must have at least 22 GB of free disk space.”
