Hacker News Reader

VRAM, not RAM. I wish it was light enough for iGPUs too

I ran the 26B model on my i5 which has no discrete graphics card. It ran about 7 tokens/sec, and appreared to be a very capable model.

Interesting, thanks for the feedback!

I've tested the e4B with the iGPU and it works incredibly good locally. I'll try this model as soon as it is available on ollama