Strix Halo GuideIndependent · measured · reproducible

AMD Strix Halo · Ryzen AI Max · Models, setup and buying

Choose a model for your RAM and get local AI running on your AMD Strix Halo PC.

Find a model to try for the memory you have, from 32GB to 192GB. Follow the step-by-step setup for a 128GB Ubuntu system or the manual route for other sizes, and get help where a step fails. Every choice shows whether it was tested here, reported by others or not yet tested.

  • Exact commands and versions
  • Positive and negative results
  • No pay-to-rank hardware

The short answer

Start simple. Move to advanced routes only when the workload needs them.

The step-by-step start route is written for a 128GB Strix Halo system on Ubuntu 24.04. Other memory sizes and Windows have their own checks and limits, so choose your route before you copy memory or boot settings.

Beginner

Start with one local chat

Follow the Linux setup, check a known answer, then look at where Ollama reports the model's memory. The guide shows which components were tested and which fresh-install and recovery checks are still open.

Read the setup and its limits
Current model

Qwen3.8 27B

20.42 generation t/s with Ollama-default MTP drafting (not a no-draft result) and exact local retrieval through 50,059 prompt tokens on the measured official route. Matched control: Ollama 0.32.15, 2026-09-26: 12.89 t/s without drafting versus 22.71 t/s with default MTP (draft_num_predict 4).

Understand every route
Buyer

Choose memory before branding

Choose memory using the exact model artifact and required context. Compare complete configurations and dated regional offers; advertised capacity is not a local qualification.

Compare the systems

Evidence, not a collage

Useful results with their claim type still attached.

These numbers answer different questions. Direct llama-bench, an Ollama API result and a low-bit capacity pass are not interchangeable leaderboard rows.

Official Qwen3.8 buyer route

Chat, image, tools, thinking and medium context.

Nine warm Ollama 0.32.13 repeats measured 292.49 prompt t/s and 20.42 generation t/s. Exact retrieval passed through 50,059 prompt tokens. These runs used Ollama-default MTP drafting; matched control: Ollama 0.32.15, 2026-09-26: 12.89 t/s without drafting versus 22.71 t/s with default MTP (draft_num_predict 4).

API evidence on one 128GB Beelink; not direct llama-bench and not a 262K local claim.

See the context boundary
20.42t/sofficial Q4_K_M, default MTP
warm generation mean

Balanced coding

96.76 t/s

Qwen3-Coder 30B-A3B

Direct llama-bench with the balanced UD-Q4_K_XL Vulkan/RADV route.

Inspect the run

Large-model capacity

284.33B

Ordinary GGUF on one box

A 90.86GB low-bit DeepSeek V4 Flash artifact loaded, generated at 13.27 tg128 and passed a narrow correctness smoke.

Capacity proof—not a broad model-quality recommendation.

Inspect the evidence

Community coverage

More than one chassis and one maintainer.

First-party Beelink measurements remain separate from community results on systems including Beelink, GMKtec, Corsair, Nimo and Minisforum. Framework is a buyer comparison, not a described owner-system reproduction in this evidence map. The count covers 10 described owner systems plus 3 external sources: 13 systems or independent sources, not 13 matched reproductions. Failures and limitations stay visible.

13
systems or independent sources
10
benchmark contributors

Why this guide exists

The differentiator is not another peak number. It is the route from purchase to repeatable result.

01

Buyers

Know what memory size, system and runtime fit the intended model.

02

Owners

Reach a working baseline without stitching together contradictory forum advice.

03

Reviewers

Use exact versions, commands, caveats and raw artifacts.

04

Vendors

See which firmware, driver and documentation gaps still block adoption.

Keep the evidence useful

Found a different result on your Strix Halo system?

Contributors keep public credit for their system, software stack, measurements and caveats. A correction or negative result is as useful as a faster row.

Independent by design.

Independent of AMD and hardware vendors. Strix Halo Guide is a project by Kleine Koe (JVD Netwerk) in the Netherlands. It is not affiliated with, endorsed by, or an official publication of AMD or any OEM. Product names identify relevant or tested hardware—not partnerships.

Read the disclosure policy