Qwen3.8 27B at 256K: 50 TPS on a 24 GB GPU | Michał Piszczek
piszczek.pl
How I built a 5.01 BPW Qwen3.8 27B hybrid, filled its 256K context, tuned MTP and gained 21.97% with a measured llama.cpp runtime.
  • AsafumEnglish
    2·
    6 days ago

    As someone who just made the mistake of buying an Intel arc b70 all I can do is read and cry with my 0 context 10 t/s worthless garbage. :(