top | item 46974214

(no title)

vkaufmann | 19 days ago

GPT-OSS-120B runs like hell on my DGX Spark

discuss

order

embedding-shape|19 days ago

The MXFP4 variant I suppose? My setup (RTX Pro 6000) does around ~140 tok/s with llama.cpp, around 160 tok/s with vLLM.

vkaufmann|19 days ago

yep MXFP4 really fast :D