#strix-halo

The latest AI and developer news about #strix-halo, with the original source and a short summary.

libstrix Docker image enables Qwen3.8-Flash-Next on Strix Halo with 40+ tok/s decode

New Docker image bundles vLLM fork and ROCm 10.1 for Strix Halo, achieving 40-43 tok/s decode for Qwen3.8-Flash-Next.

·2·2026.10.31 06:00

libstrix plugin delivers up to 31x faster prefill for vLLM on Strix Halo APUs

The open-source plugin targets gfx1151 hardware to optimize MoE GEMMs and attention layers, reducing 256k prompt prefill time from over two hours to four minutes.

·1·2026.10.29 15:00

ROCm 7.13, Strix Halo Optimizations

ROCm 7.13 nightly adds Strix Halo optimizations and open-sources ROCprof.

·2026.05.18 00:00

Luce DFlash Support for AMD Strix Halo

Luce DFlash/PFlash boosts LLM inference performance by up to more than 3x on the AMD Strix Halo iGPU.

·2·2026.05.13 03:00

Strix Halo 128B Benchmark

A Strix Halo 128GB system ran Mistral Medium 128B at 1.57 tokens per second.

·2026.05.04 09:00

Mistral Benchmark on Strix Halo

On AMD Strix Halo, Mistral Medium 3.5 took about 2 hours to process a 48k prompt.

·2·2026.05.04 03:00

ROCm Turnabout?

Comparing Vulkan and ROCm on a Strix Halo 128GB machine showed the winner varying by model.

·2026.04.17 07:00