AMD Loom boosts Qwen3.5-4B prefill up to 2.6x on Radeon 890M laptops

jfowers_amd · reddit · 2026-11-19

An AMD engineer shares benchmarks of the new Loom inference stack optimizing Qwen3.5-4B (Q4KM) on a Radeon 890M (gfx1150) laptop: up to 2.6x prefill speedup (1041 tok/s at 16k) and 1.18-1.47x decode gains over llama.cpp ROCm/Vulkan, with 4 concurrent streams at 82.3 tok/s. Model cards and quick-start guides cover gfx1150, Strix Halo and RDNA4 GPUs.

Original post →

More from Infra

Infra channel →