DeepSeek V4 Flash 284B on AMD Strix Halo (gfx1151) — up to 32 tok/s decode & ~250 tok/s prefill via ROCmFPX, DSpark & ROCm 7.2
Heterogeneous Compute Cascade (HCC) — distributed 400B-parameter MoE inference across dual AMD Ryzen AI MAX+ 395 'Strix Halo' workstations via USB4.
Converter fixes, NVFP4 findings & Strix Halo ROCmFP4 benchmarks for NVIDIA Nemotron 3.5 Lightning 30B-A3B