Two NIMO lab reference tests show how model choice, quantization and inference settings affect local throughput. Rates are in tokens/s.
| Engine / backend | Device | MTP | 16K prefill | Short decode | 16K decode |
|---|---|---|---|---|---|
| llama.cpp / ROCm-HIP | GPU | Off | 568.06 | 23.75 | 22.12 |
| Gufo / ROCm-HIP | GPU | Off | 1,888.04 | 35.03 | 34.07 |
| Gufo / ROCm-HIP + XRT | GPU + NPU | Off | 2,080.23 | 35.03 | 34.24 |
| Gufo / ROCm-HIP | GPU | On | 1,812.61 | 60.94 | 53.67 |
| Gufo / ROCm-HIP + XRT | GPU + NPU | On | 2,002.25 | 60.94 | 99.04* |
Ryzen™ AI Max+ 395, Radeon™ 8060S, 128GB-class shared memory and NimoOS. 16,384 input tokens, 256 output tokens, three measured runs after warm-up. *99.04 tokens/s reflects a sample with different output tokens and draft acceptance. Engine timing methods differ.
| Model / quantization | Engine / backend | MTP | Decode |
|---|---|---|---|
| Qwen3.6-35B-A3B · Q4_K_M | llama.cpp / ROCm-HIP | Off | 48.24 |
| Qwen3.6-35B-A3B · Q4_K_M | llama.cpp / ROCm-HIP | On | 61.00 |
| chadrock-35B-ace-saber · ROCmFP4 | ROCmFPX / ROCm-HIP 7.2.4 | On | 92.20 |
| chadrock-35B-ace-saber · ROCmFP4 | ROCmFPX / Vulkan | On | 111.44 |
NIMO Mini PC lab reference with Ryzen™ AI Max+ PRO 395, 128GB LPDDR5X and NimoOS. 32K context configuration and f16 KV cache. The 2.31× result compares the listed ROCmFP4/Vulkan/MTP configuration with the listed Q4_K_M baseline. These results characterize the test systems.
Work without limits. Up to 128GB of ultra-fast 8000MHz 256-bit LPDDR5X memory keeps large projects, real-time previews, and parallel tasks running smoothly—no stutters, no waiting, even under sustained load.

Room to grow. Dual M.2 PCIe 4.0 SSD slots deliver fast load times today and easy expansion tomorrow, keeping large files, game libraries, and creative assets instantly accessible.

See more at once. HDMI 2.1 and DisplayPort 1.4 support high-resolution, high-refresh displays—up to 8K at 120 Hz—making complex timelines, multi-window work, and immersive visuals feel effortless.

A smaller footprint, same performance. Compared with our 4L version using the same AMD AI MAX+ 395, this new compact design measures 2.76 × 7.48 × 7.87 inches instead of 3.8 × 7.4 × 9.7 inches, freeing up desk space while delivering the same class-leading power.

More ports. Less clutter. With 13 built-in connections, including USB4, HDMI, DisplayPort, and 2.5G Ethernet, displays, storage, and peripherals connect directly—no docks, no adapters, no compromise.

Power, happily compact.


The AMD AI MAX+ 395 defines a new class of performance for compact systems. With a 16-core, 32-thread architecture and boost speeds up to 5.10 GHz, it delivers exceptional CPU and graphics performance with fast, unified memory — enabling fluid multitasking, advanced creation, and responsive computing without compromise.
Designed with an integrated AI engine delivering up to 126 TOPS of total AI performance, including a dedicated 50 TOPS NPU, the NIMO AI MAX+ 395 Mini PC is built for the next generation of personal AI computing. Modern workloads, complex AI tasks, and real-time processing remain fast, responsive, and efficient in a remarkably clean and compact setup, making it perfectly suited for running advanced local AI tools like OpenClaw. Deploy intelligent agents, automate workflows, and process tasks directly on your device with powerful performance while keeping your data fully under your control — a smarter, faster, and more private way to experience AI.


The integrated graphics(8060S) of the AMD Ryzen AI Max+ 395 is currently the world’s most powerful iGPU, featuring 40 CUs clocked up to 2.9 GHz and leveraging the advanced RDNA 3.5 architecture. This powerful iGPU outperforms an RTX 4060 GPU in rendering benchmarks, delivering fluid FHD performance even in demanding creative and gaming workloads.
From studying and everyday office work to photo editing, coding, design, music production, video creation, and gaming — this Mini PC is built to handle it all smoothly. One compact system, ready for whatever the day demands.


Smooth multitasking keeps daily work flowing without slowdowns. Large files, multiple apps, and long workdays stay responsive in a clean, clutter-free workspace.


Immersive gameplay runs fluidly without the need for a bulky setup. High frame rates and stable performance make after-hours entertainment feel effortless.


Fast builds and real-time testing stay uninterrupted, even with heavy workloads running in parallel. A compact system that keeps focus on code, not wait times.


Creative tools stay responsive from concept to final export. Complex visuals, layered projects, and detailed renders move smoothly without breaking momentum.


Editing timelines stay smooth from rough cut to final export. High-resolution footage, effects, and color grading flow without dropped frames or long render waits.
En Nimo, creemos que todos merecen acceso a tecnología de alta calidad. Por eso ofrecemos laptops reacondicionadas certificadas por Nimo, restauradas minuciosamente y sometidas a pruebas rigurosas para cumplir con nuestros exigentes estándares. Sabemos que la condición estética es importante, por lo que hemos adoptado un sistema de clasificación claro y transparente.
Conoce los grados de reacondicionamiento de Nimo
Al elegir una laptop reacondicionada certificada por Nimo, eliges calidad, valor y sostenibilidad. Utilizamos un sistema de clasificación sencillo para ayudarte a comprender la condición estética de tu dispositivo:
¿Qué distingue a los productos reacondicionados certificados por Nimo?
¿Por qué elegir productos reacondicionados certificados por Nimo?
Compra con confianza
Sabemos que comprar una laptop reacondicionada puede generar dudas. Por eso, nos comprometemos a proporcionarte toda la información que necesitas para tomar una decisión informada. Si tienes preguntas, nuestro amable equipo de atención al cliente siempre está 1 aquí para ayudarte.
¿Tienes una cuenta?
Inicia sesión para finalizar tus compras con mayor rapidez.