Optimization Guide On Device Model - Version: 0.0.0.0

188 views
Skip to first unread message

Oyetunde Dotun

unread,
Oct 13, 2025, 9:40:13 AM10/13/25
to Chrome Built-in AI Early Preview Program Discussions
Hi everyone, it's so sad I might not be able to paarticipate in the ongoing hackathon, given that I've got a unique and great idea.

The optimization guide on device model failed to download, it just stuck at Checking for Status, I've tried all possible means even with the discussions on chromium but all has failed.

I have enabled all the flags:
Enables optimization guide on device, Prompt API for Gemini Nano, Prompt API for Gemini Nano with Multimodal Input.

My chrome version: 142.0.7444.23 (Official Build) beta (64-bit) (cohort: Beta)

And I met all the foundational model criteria

Is there anything else I am missing? My disk space is around 33Gb free space.

Nikita Malyschkin

unread,
Oct 13, 2025, 9:53:38 AM10/13/25
to Chrome Built-in AI Early Preview Program Discussions, Oyetunde Dotun
This is not a solution but might be a work around:

You can use Ollama or localAI and use the locally exposed API to develop your idea.
I know it's not ideal but you could get started on your project and if you mimic the chrome AI interface you could even make sure that your application works on other machines.
Let me know if you want to try it out and could use some pointers :)

Oyetunde Dotun

unread,
Oct 13, 2025, 9:56:25 AM10/13/25
to Chrome Built-in AI Early Preview Program Discussions, Nikita Malyschkin, Oyetunde Dotun
Yeah sure, I'd want to try that out.

Nikita Malyschkin

unread,
Oct 13, 2025, 10:04:09 AM10/13/25
to Chrome Built-in AI Early Preview Program Discussions, Oyetunde Dotun, Nikita Malyschkin
https://github.com/ollama/ollama?tab=readme-ov-file#rest-api
https://localai.io/features/text-generation/

Be aware that these tools have different interfaces and limitations than the built in API, generally being more powerful.
The Ollama interface resembles the chrome API best and is rather easy to get started with.

Oyetunde Dotun

unread,
Oct 13, 2025, 10:13:56 AM10/13/25
to Chrome Built-in AI Early Preview Program Discussions, Nikita Malyschkin, Oyetunde Dotun
But aren't these apis a different model other than the ones provided by google and needed for this hackathon?

Nikita Malyschkin

unread,
Oct 13, 2025, 10:18:17 AM10/13/25
to Chrome Built-in AI Early Preview Program Discussions, Oyetunde Dotun, Nikita Malyschkin
You can select the model in the API. If I'm not mistaken the model used in chrome is gemma3n, which is available with Ollama.

Vishal Verma

unread,
Oct 13, 2025, 10:32:09 AM10/13/25
to Nikita Malyschkin, Chrome Built-in AI Early Preview Program Discussions, Oyetunde Dotun
There's also the option to manually load the model, which I haven't tried. chrome://on-device-internals or something 


--
You received this message because you are subscribed to the Google Groups "Chrome Built-in AI Early Preview Program Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chrome-ai-dev-previe...@chromium.org.
To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/chrome-ai-dev-preview-discuss/96d0909d-14e4-4e6d-9a1a-00b1b6cdf48en%40chromium.org.

Thomas Steiner

unread,
Oct 13, 2025, 11:27:41 AM10/13/25
to Vishal Verma, Nikita Malyschkin, Chrome Built-in AI Early Preview Program Discussions, Oyetunde Dotun
I've also mentioned streamed browsers a couple of times now when people's devices don't meet the requirements bar. Search this group's archive for browserstack



--
Thomas Steiner, PhD—Developer Relations Engineer (blog.tomayac.comtoot.cafe/@tomayac)

Google Spain, S.L.U.
Torre Picasso, Pl. Pablo Ruiz Picasso, 1, Tetuán, 28020 Madrid, Spain

CIF: B63272603
Inscrita en el Registro Mercantil de Madrid, sección 8, Hoja M­-435397 Tomo 24227 Folio 25

----- BEGIN PGP SIGNATURE -----
Version: GnuPG v2.4.8 (GNU/Linux)

iFy0uwAntT0bE3xtRa5AfeCheCkthAtTh3reSabiGbl0ck
0fjumBl3DCharaCTersAttH3b0ttom.xKcd.cOm/1181.
----- END PGP SIGNATURE -----
Reply all
Reply to author
Forward
0 new messages