Qwen3.6 27B
on 2× AMD MI50 32GB · llama.cpp
May 19, 2026
Summary
User benchmarks MTP KV cache quantization at Q8_0 against no quantization on Qwen3.6-27B-Q8_0 with llama.cpp.
Setup uses tensor parallelism on 2xMI50 32GB.
Wall time difference is negligible at roughly 0.14s with the quantized draft KV cache. Aggregate accept rate is 0.735-0.741.