X (Twitter)

kyuz0.github.io

August 26, 2026

AMD Strix Halo — Backend Benchmarks (Grid View)

Backend benchmarks for AMD Strix Halo measuring tokens-per-second throughput across 17 LLM configurations ranging from 6.7B to 228.7B parameters. Test...

youtube.com

September 24, 2025

GLM 4.5-Air-106B and Qwen3-235B on AMD “Strix Halo” AI Ryzen MAX+ 395 (HP Z2 G1a Mini Workstation)

In this video, I show how to run large language models like Qwen3-235B in Q3_K_XL and GLM 4.5-Air-106B in Q6_K_XL on any AMD Ryzen AI MAX “Strix Halo”...

github.com

August 19, 2026

GitHub - julianmb/q38rocm: 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.

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

~/bookmarks

Running 122B-Parameter LLMs Locally on AMD Strix Halo for OpenClaw:...

Running 122B-Parameter LLMs Locally on AMD Strix Halo for OpenClaw:...

~/bookmarks

Running 122B-Parameter LLMs Locally on AMD Strix Halo for OpenClaw:...

Discover More

AMD Strix Halo — Backend Benchmarks (Grid View)

GLM 4.5-Air-106B and Qwen3-235B on AMD “Strix Halo” AI Ryzen MAX+ 395 (HP Z2 G1a Mini Workstation)

GitHub - julianmb/q38rocm: 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.

Similar Content

Running 122B-Parameter LLMs Locally on AMD Strix Halo for OpenClaw:...

Discover More

AMD Strix Halo — Backend Benchmarks (Grid View)

GLM 4.5-Air-106B and Qwen3-235B on AMD “Strix Halo” AI Ryzen MAX+ 395 (HP Z2 G1a Mini Workstation)

GitHub - julianmb/q38rocm: 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.