Qwen3.8 27B
on AMD RX 7900 XTX · llama.cpp · 32,768 ctx
Use cases
codingcreative-writing
Summary
User reports Qwen3.8 27B at 40 t/s on an RX 7900 XTX 24 GB, with 32k context.
Setup is llama.cpp with ROCm 7.2 Windows binaries provided by AMD, Q4_K_M quant, 32 GB system RAM.
A Qwen3 A3B 30B Instruct Q4_K_M run at 32k context reached 140 t/s. The user plans to upgrade RAM to 48 GB and uses Tailscale to reach the machine from three devices.