V100 32 GB VRAM, 32GB RAM

llama-server \
      --host 0.0.0.0 \
      --port 8080 \
      --model unsloth-Gemma-4-31B-it-qat/gemma-4-31B-it-qat-UD-Q4_K_XL.gguf \
      --mmproj unsloth-Gemma-4-31B-it-qat/mmproj-F16.gguf \
      --model-draft unsloth-Gemma-4-31B-it-qat/mtp-gemma-4-31B-it.gguf \
      --alias 'gemma4' \
      --jinja \
      --gpu-layers all \
      --temp 1.0 \
      --top-p 0.95 \
      --top-k 64 \
      --ctx-size 196608 \
      -ctk q8_0 -ctv q8_0 \
      -b 512 -ub 512 \
      --spec-type draft-mtp \
      --spec-draft-n-max 3 \
      --parallel 2 \
      --ctx-checkpoints 4 \
      --checkpoint-min-step 32768 \
      --cache-ram 22528 \
      --cache-idle-slots