Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →You can run local language models on a Ryzen AI Max workstation with Ollama on a documented Ubuntu setup, or with llama.cpp using a compatible backend. Windows has a separate AMD-documented LM Studio and Vulkan route. The right setup depends on your exact system, driver, model, quantization, and memory allocation: Ryzen AI Max machines do not all have the same memory configuration, and a model’s weights are only part of its memory use.
How do I run an LLM locally on Ryzen AI Max?
For the simplest documented Linux path, follow AMD’s example for a Ryzen AI Max+ 395 with 128GB of unified system memory, Ubuntu 24.04 LTS, ROCm 7.2.1, and Ollama 0.20.x. Treat those versions as a tested example, not a universal current recipe: before installing, check AMD’s current compatibility documentation for your specific processor, operating system, and driver/runtime combination.
As an Amazon Associate I earn from qualifying purchases.
- Confirm the system and software match. AMD’s walkthrough uses a Ryzen AI Max+ 395 system with 128GB unified memory and configures 64GB as GPU-accessible memory. Check your machine’s memory configuration and the current ROCm compatibility information before proceeding.
- Install Ollama for your Ubuntu environment. Use the installation method and version appropriate to the AMD walkthrough and your supported ROCm setup. Do not assume that an Ollama installation by itself enables GPU inference.
- Download the example model. In a terminal, run
ollama pull qwen3.5:35b. - Start a chat with it. Run
ollama run qwen3.5:35b. - Inspect placement while it is loaded. In another terminal, run
ollama ps. AMD’s all-GPU placement result applies to its Qwen3.5 35B demonstration under the specified configuration; it does not mean every model will fit entirely in GPU-accessible memory.
The exact commands are from AMD’s May 25, 2026 Ollama walkthrough. If your device, Ubuntu release, ROCm version, or Ollama version differs, first verify that combination rather than copying the version numbers blindly.
Can I use Ollama on Ryzen AI Max?
Yes. AMD documents Ollama on Linux using ROCm, including the Ubuntu example above. The published walkthrough tested Qwen3.5 9B, 35B-A3B, and 122B-A10B at Q4_K_M quantization. These are AMD-reported demonstrations, not independent performance tests.
#1 Best Overall
- Built for Local AI Development: AMD Ryzen AI Halo is designed for local AI development and inference, featuring 128GB unified memory and support for up to 200B parameter models to build and run intensive AI workloads locally.
- 128GB Unified Memory: Features 128GB LPDDR5x unified memory at 8000 MT/s with 256 GB/s memory bandwidth, providing a shared memory pool across the CPU, GPU, and NPU to support larger AI models.
- AMD Ryzen AI Max+ 395 Processor: Features 16 cores, 32 threads, and Zen 5 architecture, paired with AMD Radeon 8060S integrated graphics featuring 40 RDNA 3.5 compute units and an AMD XDNA 2 NPU with up to 50 TOPS.
- Linux AI Developer Platform: Purpose-built for Linux-based AI development with full AMD ROCm software support and preloaded tools, models, and workflows optimized for local AI development.
- Compact, Connected Design: Includes a 2TB M.2 SSD, 10GbE LAN, Wi-Fi 7, Bluetooth 5.4, USB-C connectivity, and HDMI 2.1b.
For the Qwen3.5 122B-A10B example, AMD reports a 76GB model footprint. Because that exceeds the walkthrough’s configured 64GB of GPU-accessible memory, AMD loaded it with mixed placement: 61% GPU and 39% CPU. That illustrates why “model runs” and “model fits entirely in GPU memory” are different claims.
When should you use llama.cpp instead?
Choose llama.cpp when you want more direct control over GGUF models, quantization, and how many model layers are offloaded to the GPU. It is a lower-level alternative to Ollama’s simpler model-management workflow, so expect to choose and verify more settings yourself.
Rank #2
- 【Leading AI Mini Workstation】MINISFORUM AI MS-S1 Max Workstation comes with AMD Ryzen AI Max+ 395 processor, which uses AMD's latest generation Zen 5 architecture. It has 16 Cores and 32 Threads, the boost clock is up to 5.1GHz. The overall processor performance is up to 126 TOPS, and the NPU performance reaches up to 50 TOPS. AMD Ryzen AI enables improved productivity, advanced collaboration, and improved efficiency.
- 【AMD Radeon 8060S Graphics 】The MS-S1 Max Mini PC equipped with AMD Radeon 8060S Graphics which built on the new generation of RDNA 3.5 architecture AMD graphics, it brings ultra-high frame rate experiences and advanced content creation features anywhere and delivers staggering performance. It can handle all your computing and multimedia tasks efficiently.
- 【Five 8K Video Output】This MS-S1 Max Workstation comes with five video outputs, 1x HDMI (8K@60Hz), 2x USB4(40Gbps,Alt DP2.0,PD out 15W) and 2x USB4 V2(80Gbps,Alt DP2.0,PD out 15W) Outputs, which support multiple monitors display at the same time and provide a larger and wider filed of view and improve your work efficiency. It is used in fields that require high-performance computing and graphics processing, including digital signage and securities trading, as well as work that uses CAD, such as engineering design, scientific calculations, animation production, and post-production for movies and television
- 【 Fast and Stable Wire & Wireless Speed】It comes with Two 10G Lan Ports for wired connection and and Wi-Fi 7 / BT5.4 for wireless connection, which increased the network speed greatly and expand its functions and improved performance of computer to a large extent and allows you to use more networks such as software routers (OpenWRT / DD-WRT / Tomato etc.), firewalls, NAT, network isolation etc.
- 【Large Storage & Flexible Expandability】This Workstation equipped with 64GB LPDDR5-8000MHz + 2TB M.2 2280 PCIe4.0 SSD. There is another PCIe4.0 SSD slot available for up to 8TB, these SSD slots are compatible with RAID0 and RAID1, you can store movies, videos, photos, important files easily. What’s more, it also comes with 1x standard PCIex16 slot(PCIe4.0x4) inside.
- Check support for the exact APU and operating system. AMD’s llama.cpp on ROCm documentation provides Linux and Windows instructions, but support depends on the specific hardware and software environment. An integrated GPU showing shared system memory does not prove that it is a supported or usable ROCm device.
- Install the matching driver and runtime. Follow the instructions for the operating system and supported device, including Linux package and group prerequisites where applicable. Avoid mixing instructions for different distributions or runtime versions.
- Select a suitable GGUF model and quantization. Start with a smaller quantized model and a moderate context length, then increase either after checking memory use and responsiveness.
- Set GPU offload deliberately and verify the selected device. AMD documents
llama-benchwith a GGUF model and-ngl 999for benchmark reproduction. That is not a universal setting for every model or machine; use model-appropriate settings and confirm which device the runtime actually selected.
AMD’s ROCm 7.2.1 limitations page specifically cautions that some LLM workloads may perform below expectations on Ryzen AI Max+ 395 processors. Validate your own model and software combination rather than assuming a supported launch guarantees the responsiveness you want.
What is the documented Windows setup?
Windows is a distinct path from the Ubuntu ROCm walkthrough. AMD’s Windows example uses LM Studio with llama.cpp and Vulkan, AMD Adrenalin Edition 25.8.1, and Variable Graphics Memory (VGM). It is not evidence that Windows ROCm support is interchangeable with Linux ROCm, or that every Ryzen AI Max system supports the same features.
Rank #3
- Built for Local AI Development: AMD Ryzen AI Halo is designed for local AI development and inference, featuring 128GB unified memory and support for up to 200B parameter models to build and run intensive AI workloads locally.
- 128GB Unified Memory: Features 128GB LPDDR5x unified memory at 8000 MT/s with 256 GB/s memory bandwidth, providing a shared memory pool across the CPU, GPU, and NPU to support larger AI models.
- AMD Ryzen AI Max+ 395 Processor: Features 16 cores, 32 threads, and Zen 5 architecture, paired with AMD Radeon 8060S integrated graphics featuring 40 RDNA 3.5 compute units and an AMD XDNA 2 NPU with up to 50 TOPS.
- Windows 11 Pro AI Developer Platform: Built for AI development on Windows 11 Pro with AMD ROCm software support and access to tools, models, and workflows for local AI development.
- Compact, Connected Design: Includes a 2TB M.2 SSD, 10GbE LAN, Wi-Fi 7, Bluetooth 5.4, USB-C connectivity, and HDMI 2.1b.
In that 2025 example, AMD describes up to 96GB of VGM on a Ryzen AI Max+ 395 system with 128GB of total memory. The amount is configuration- and driver-specific; it is not a promise that any Ryzen AI Max machine can allocate 96GB to graphics or that all of it is available to a model after other memory use.
AMD also reports a Llama 4 Scout run with 109B total parameters and 17B active parameters, up to 15 tokens per second, and a 256,000-token context. The context result is reported with Flash Attention enabled and a Q8 KV cache on the stated driver. These are vendor-reported results for that configuration, not independent measurements or a general speed guarantee. Although only 17B parameters are active, the model’s weights still need to be held in memory.
Rank #4
- 【High-Performance APU】The MS-S1 MAX features an AMD Ryzen AI Max+ 395 APU, integrating a Zen 5 architecture CPU (up to 5.1GHz, 16C/32T, 64M L3 Cache), an RDNA 3.5 GPU, and an NPU (50 TOPS). The total system output is 126 TOPS. It provides powerful parallel computing capabilities for demanding AI workflows. It is ideal for running local LLMs, multimodal models, and computationally intensive tasks
- 【128GB UMA Memory】Equipped with up to 128GB of LPDDR5x-8000MT/s unified memory, it enables the CPU and GPU to access a shared, high-bandwidth memory pool with extremely low latency. Ideal for large-scale AI inference, 3D workloads, and complex timelines in video editing. It eliminates traditional VRAM bottlenecks, ensuring smoother data transfer during high-intensity computations. The UMA design maximizes performance stability under high loads
- 【Flexible Expansion】The MS-S1 MAX features USB4 V2 (up to 80Gbps), dual 10GbE LAN, HDMI 2.1 (up to 8K60), a full-length PCIe x16 expansion slot, and dual M.2 slots supporting up to 16TB RAID 0/1. Wi-Fi 7 provides stronger signal coverage and a more stable wireless experience. The slide-out design facilitates upgrades and maintenance. It easily adapts to personal, studio, or rack-mount enterprise environments
- 【High-Efficiency Cooling System】Utilizing an aerospace-grade aluminum alloy chassis, copper base plate, six heat pipes, dual turbine fans, and advanced PCM thermal conductive material, it maintains stable cooling performance even under continuous load. This system supports 130W continuous power and 160W peak power operation, with a built-in 320W power supply. It boasts multiple global certifications including CCC, FCC, UL, CE, and UKCA, ensuring stable and reliable operation in various environments
- 【Cluster Design】Two MS-S1 MAX units can be configured as a dual-unit cluster to run a large 235B Q4 model locally, achieving an output speed of 10.87 tok/s. Supporting 2U rack deployment, multiple MS-S1 MAX units can be cascaded into a distributed cluster to create a high-efficiency AI computing center. A cluster of four MS-S1 MAX units successfully ran a DeepSeek-R1 671B Q4 large model. A reserved cluster power-on interface allows for unified start-up and shutdown
See AMD’s Windows LM Studio and Ryzen AI Max example for its specific settings. Check the current LM Studio, driver, Vulkan, and system compatibility information before reproducing it.
Recommended Free Tools
How much memory do local models need on Ryzen AI Max?
There is no single minimum that applies to every model or workload. A model’s parameter count is only a rough guide: quantization changes the weight footprint, while longer context and the KV cache consume additional memory. The operating system, other running applications, GPU-accessible allocation, and CPU/GPU offload also affect whether a workload fits and how it feels to use.
Best Value
- UNOPENED RETAIL PACKAGING, sold as configured by Lenovo. Includes one year of Courier or Carry-in Lenovo Warranty. Add up to 5 years of Lenovo Premier Onsite Support Plus when you register your computer with Lenovo.
- The 14″ Lenovo ThinkPad P14s Gen 7 is an ultra-portable mobile workstation for on-the-go professionals. Featuring the robust AMD Ryzen AI 7 PRO 450 Processor, integrated Radeon graphics, and scalable memory, this Copilot+ PC offers exceptional AI performance for handling data-heavy tasks.
- Plenty of ports: 1x USB-A (USB 5Gbps / USB 3.2 Gen 1); 1x USB-A (USB 5Gbps / USB 3.2 Gen 1), Always On; 2x Thunderbolt 4, with USB PD 3.0; 1x HDMI 2.1, up to 4K/60Hz; 1x Headphone / microphone combo jack (3.5mm); 1x Ethernet (GbE RJ-45); and 1x Kensington Nano Security Slot.
- Experience outstanding visual clarity on the 14" WUXGA (1920 x 1200) IPS touchscreen display. Featuring an anti-glare finish, 500 nits of brightness, and 100% sRGB color accuracy, this low-power display delivers vibrant and crisp visuals for all your professional needs.
- Equipped with 32GB of lightning-fast DDR5-5600MT/s memory and a spacious 1TB M.2 2280 PCIe Gen4 TLC Opal SSD, providing rapid performance and ample, high-speed storage for all your professional applications and data.
- Total unified memory is not the same as GPU allocation. AMD’s Linux example has 128GB of total system memory and configures 64GB as GPU-accessible. Its Windows example describes up to 96GB of VGM on a 128GB system using the stated driver. These are different operating-system and software configurations, not interchangeable guarantees.
- Quantized weights reduce, but do not eliminate, memory needs. AMD’s Linux Qwen examples use Q4_K_M; context and runtime overhead still matter beyond the reported model footprint.
- Long context can change the fit. A large context can require substantial KV-cache memory. AMD’s Windows 256,000-token result is tied to Flash Attention and Q8 KV-cache settings, not a default expectation for other models or runtimes.
- CPU offload can make a larger model load, with trade-offs. AMD’s 122B example used both GPU and CPU placement. A model that loads this way may not respond like one fully resident on the GPU.
A practical approach is to begin with a smaller quantized model and a shorter context. Confirm that it loads, check GPU placement with the runtime’s reporting tools, and assess responsiveness before increasing model size or context.
Which Ryzen AI Max workstation should you use?
Choose by the installed memory configuration, operating system, cooling and form factor, and verified support for the inference backend—not just the Ryzen AI Max name. AMD lists Framework Desktop, ASUS ROG Flow Z13, HP ZBook Ultra G1a, Corsair AI Workstation 300, and HP Z2 Mini G1a among systems available in 128GB Ryzen AI Max+ configurations. Availability and exact configurations vary; confirm the specification of the system you are considering.
AMD’s examples are useful evidence that these setups can run large local models, but they are vendor demonstrations rather than independent comparisons. Actual compatibility and performance depend on the particular machine, software versions, model, quantization, context, and offload settings.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
