Guided setup scripts to run Qwen3.8-Flash-Next (180B, 6B active) on a single AMD Strix Halo (128 GB RAM, 130 GiB disk) using the halogen-flash-server container.
Targets Fedora, Ubuntu, and Arch; exposes an OpenAI-compatible API with configurable port, vision toggle, and parallel slots. Sustained decode of 42–45 tok/s; integrates with coding agents such as PI and OpenCode.
Qwen 3.8 27B ROCmFP4 on AMD Strix Halo (Ryzen AI Max+ 395). Up to 36 tok/s via MTP Speculation, TurboQuant & Mesa RADV Wave64. - julianmb/q38rocm
High-performance Lemonade alternative for AMD Strix Halo & Radeon — latest MoE models (Ling-3.0-Flash, Qwen 3.8 Flash Next, Ornith 1.5), bleeding-...
Strix Halo inference engine. Qwen Flash Next Q4_K_XL: 1,628.52pp, 59.41tg single user, 157.22 tok/s 8 users; Qwen27B Q4_K_XL: 656.33pp, 70.56tg tok/s ...
