qwen3.6-27b-fp8-mtp-vllm

vllmtorch@official

Qwen3.6-27B-FP8 — native FP8 format

Quick Info

Runtimevllm (torch)

Tensor Parallel1

Nodes1+

Dtypefp8

Containerghcr.io/spark-arena/dgx-vllm-eugr-nightly:latest

Tensor Parallel (--tp)GPU Memory (--gpu-mem)Port (--port)Served Model Name

sparkrun run @official/qwen3.6-27b-fp8-mtp-vllm