you are viewing a single comment's thread
view the rest of the comments
[–] 2 points 6 days ago

Local inference on AMD is getting crazy these days. 90-100tps decode for Qwen 3.8 Flash Next on a single R9700.

  • source
  • parent