Issues

• Eval bug: -sm tensor makes model loading take forever before eventually crashing

  mxxm-t/mx-llama.cpp#19 · Filter by author Haron123 opened on Oct 2, 2026

• Eval bug: MiMo v2.6 Flash outputs gibberish

  mxxm-t/mx-llama.cpp#18 · Filter by author Te-eMster opened on Sep 25, 2026

• Working 4x MI50 (gfx906) recipe for Qwen3.8-Flash-Next + note: tensor-split needs -fa on

  mxxm-t/mx-llama.cpp#11 · Filter by author JackDanger opened on Sep 5, 2026
  2

• Small-batch matmul cost curve on 2x MI50 (Q8_0 repack, tensor split)

  mxxm-t/mx-llama.cpp#10 · Filter by author JCraigWasTaken opened on Sep 4, 2026

• Eval bug: 2x MI50 tensor split aborts in RCCL on the first AllReduce once peer access works, no fallback to butterfly

  stale
  mxxm-t/mx-llama.cpp#7 · Filter by author JCraigWasTaken opened on Sep 2, 2026
  4