Post

Chrome shipped an LLM Prompt API to the web platform. At Mozilla, we oppose this API. Here's why:

@paul.kinlan.me

@jakearchibald.com - "In Chrome when you create a LanguageModel it will either use the Google Cloud API or download a 4GB model to run locally". We don't use the Google Cloud API. If the model isn't available we will kick off a download, but the API just doesn't work without the modal locally.

Firefox for Web Developers

@webdevs.firefox.com

18 May

Interesting! When if I uninstall the model, and chrome://on-device-internals/ confirms the folder size has gone from 4gb to 0MiB, the prompt API still works. Including in Incognito. What's going on here?

fry69

@fry69.dev

Even after full browser restart? Otherwise the model could still be loaded in memory?

18/5/26, 3:19 pm0 reposts3 likes

Firefox for Web Developers

@webdevs.firefox.com

18 May

Yeah it seems like Chrome misreports that the model is deleted when it isn't.

fry69

@fry69.dev

18 May

If the model continues to work even after deleting it and a clean browser restart, that is concerning indeed 🤔

Firefox for Web Developers

@webdevs.firefox.com

18 May

After a reset, it immediately redownloads, even if the API isn't called.

fry69

@fry69.dev

18 May

Ah 💡 Side note: I noticed that other Google apps now also seem to be more forcefully, circumventing user interactions. Eg the native Gemini App creates "start on login" entry without announcing it or a way to remove it. When this entry gets manually removed, Gemini App puts it automatically back.

Tom Hicks

@hicksyfern.bsky.social

18 May

Is this good? Seriously though, your objections are hard to argue with.