@bytebot
I am very excited by Gemma 4.
You can even ask it to describe what it sees, i.e. a video tool call.
Lowest I have been able to test it on so far? Late 2019 Intel MacBook Pro (16GB RAM/1.5GB Intel UHD Graphics 630 + 4GB AMD Radeon Pro 5500M).
It works. 3.8GB stored in Chrome cache. We love transformers.js!
And we will use this for Mitsukeru LinkSentinel and other pieces of kit - more embedded models, good for privacy.
Do you have an older box to give it a try?
https://huggingface.co/spaces/webml-community/Gemma-4-WebGPU
1 reply
0 recast
8 reactions