64GB max RAM · 6 bays · 2 x 10GbE
AINAS Score 75
No verified price

Native AI NASMinisforum
Why Native AI NAS: AMD Ryzen AI Max+ 395 (16C/32T) with 50 TOPS NPU and Radeon 8060S on 64GB or 128GB LPDDR5X unified memory, and Minisforum ships a local LLM/RAG stack (MinisOpenClaw, Ollama/llama.cpp knowledge base).
Updated
Can the Minisforum N5 Max run local AI?
Yes, the Minisforum N5 Max can run local language models. It pairs AMD Ryzen AI Max+ 395 and a 50 TOPS NPU with up to 128GB of memory, which suits 70B class models at higher-quality quantization, or large mixture-of-experts models at 4-bit (Estimated).
There is no way to add a discrete GPU, so all inference runs on the CPU and basic integrated graphics.
Local LLM support is documented by the manufacturer.
The N5 Max suits buyers who want a NAS that also runs local AI models. It pairs the AMD Ryzen AI Max+ 395 (16 cores) with a 50 TOPS NPU and the Radeon 8060S on LPDDR5X unified memory, 64GB included and up to 128GB. Llama 3.3 70B fits at Q4 (Estimated), and 32B-class models fit at both Q4 and Q8 (Estimated). Minisforum includes a built-in knowledge base with RAG and citations, vendor confirmed. Storage is generous: 5 drive bays, 5 M.2 slots and 2 x 10GbE. Docker and virtual machines are supported. The AINAS Score is 71. Warranty is 2 years.
The main catch is that the memory is soldered and cannot be upgraded. The N5 Max has no PCIe slot and no OCuLink, so there is no graphics card path either. Buyers who expect to grow into larger models later are affected, because the memory you buy is the memory you keep. Llama 4 Scout 17Bx16E does not fit at Q4 (Estimated). Anyone who needs more headroom should look for a system with at least 128GB of unified memory from day one, or one with a PCIe slot or OCuLink port for a dedicated GPU. The memory has no ECC stated either.
Amazon US
64GB LPDDR5X + 128GB OS storage
Ships from the US; delivery and import charges shown at checkout
$2,399.00
Verified
In stock
Buy on AmazonAmazon US
128GB LPDDR5X + 128GB OS storage
Ships from the US; delivery and import charges shown at checkout
Price awaiting re-verification
Last verified · checked
Out of stock
Buy on AmazonProduct prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon.co.uk or Amazon.com, as applicable, at the time of purchase will apply to the purchase of this product.
Each capability shows how it was confirmed: Verified by AINAS, Vendor confirmed, Community confirmed or Unverified.
| Capability | Supported | Evidence | Note |
|---|---|---|---|
| Local LLM | Yes | Vendor confirmed | Marketed as 35B-class local model ready (Qwen3.6 35B, Gemma 4 26B); MinisOpenClaw local AI agent. |
| Ollama | Yes | Vendor confirmed | White paper: llama.cpp core runtime with Ollama optional runtime; example Ollama (Qwen 3.6 35B). |
| Open WebUI | Limited | Unverified | Own FastAPI web UI for the knowledge base; Open WebUI not documented. |
| RAG | Yes | Vendor confirmed | Built-in local knowledge base: hybrid BM25 + vector search, RAG with citations. |
| AI search | Yes | Vendor confirmed | File-level search plus AI semantic search; smart photo search. |
| GPU expansion | Limited | Unverified | No OCuLink or PCIe slot; only USB4 v2 80Gbps ports, eGPU use not documented. |
| Hardware acceleration | Yes | Vendor confirmed | 50 TOPS NPU and Radeon 8060S (AMD); Minisforum markets up to 126 TOPS platform. |
| AI photos | Yes | Vendor confirmed | Smart album and smart photo search. |
| AI video | No | Unverified | No video AI documented. |
| Surveillance AI | No | Unverified | No surveillance app documented. |
Calculated from the N5 Max's memory and each model's published size at each quantization. Rows marked Estimated are calculations, not benchmarks. Check any model.
| Model | Best quantization | Runs? | Memory needed | Expected performance | Verification |
|---|---|---|---|---|---|
| Llama 3.2 3BLlama, 3.21B | Q8 | Yes | 13GB | Not measured | Estimated |
| Llama 3.1 8BLlama, 8.03B | Q8 | Yes | 19GB | Not measured | Estimated |
| Qwen3 8BQwen, 8.19B | Q8 | Yes | 19GB | Not measured | Estimated |
| Gemma 3 12BGemma, 11.8B | Q8 | Yes | 24GB | Not measured | Estimated |
| Qwen3 14BQwen, 14.8B | Q8 | Yes | 28GB | Not measured | Estimated |
| Phi-4 14BPhi, 14.7B | Q8 | Yes | 28GB | Not measured | Estimated |
| Mistral Small 3 24BMistral, 23.6B | Q8 | Yes | 38GB | Not measured | Estimated |
| Gemma 3 27BGemma, 27B | Q8 | Yes | 44GB | Not measured | Estimated |
| Qwen3 30B A3BQwen, 30.5B | Q8 | Yes | 47GB | Not measured | Estimated |
| Qwen3 32BQwen, 32.8B | Q8 | Yes | 50GB | Not measured | Estimated |
| DeepSeek-R1 32B (Qwen distill)DeepSeek, 32.8B | Q8 | Yes | 50GB | Not measured | Estimated |
| Llama 3.3 70BLlama, 70.6B | Q4 | Yes | 60GB | Not measured | Estimated |
| Llama 4 Scout 17Bx16ELlama, 109B | None fits | No | 89GB at Q4 | Not measured | Estimated |
| Qwen2.5 72BQwen, 72.7B | Q3 | Yes | 54GB | Not measured | Estimated |
| Mistral Large 2 123BMistral, 123B | Q2 | Yes | 62GB | Not measured | Estimated |
One verdict per workload, from the N5 Max's use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
The N5 Max is a good fit for running local language models. Minisforum markets it as ready for 35B-class models (vendor confirmed), and Llama 3.3 70B fits at Q4 (Estimated) on the unified memory.
Private RAG is a good fit here. A built-in knowledge base offers hybrid BM25 and vector search with cited answers (vendor confirmed), though Open WebUI is unverified and not documented.
AI search is a good fit on the N5 Max. Vendor-confirmed file-level search sits alongside AI semantic search, so you can find documents by meaning as well as by name.
AI photo handling is a good fit. Smart albums and smart photo search are vendor confirmed, and the 5 drive bays and 5 M.2 slots give room for a large photo library.
Surveillance is a poor fit for the N5 Max. No surveillance app is documented, and that capability is unverified, so do not plan camera recording around it.
Home lab use is a good fit. Docker is supported with per-account container isolation, virtual machines are listed, and 2 x 10GbE plus 3 USB4 ports give flexible connections.
Business use is a good fit. The N5 Max offers 5 bays, 2 x 10GbE, Docker and a 2-year warranty, though there is no ECC memory for buyers who require it.
Creative AI is an OK fit. The Radeon 8060S and 50 TOPS NPU are vendor confirmed, but there is no PCIe or OCuLink path, and eGPU use over USB4 is unconfirmed.
The N5 Max ships with 64GB of LPDDR5X 8533 (soldered, 256-bit) that cannot be upgraded, so the configuration you buy is the one you keep. For local AI that matters more than anything else on the spec sheet: model size is capped by memory, not storage.
As a rule of thumb, a 4-bit model needs about 0.6GB per billion parameters plus room for context. After leaving around 8GB for the NAS operating system, 128GB leaves space for a model of roughly 200B parameters at most (Estimated).
No. The N5 Max has no PCIe slot or OCuLink port for a graphics card, and eGPU use over USB4 or Thunderbolt is not a supported NAS setup. Inference runs on the CPU, with AMD Radeon 8060S handling media tasks, so keep to small models.
Its 50 TOPS NPU is used by the vendor's own AI features. Open-source LLM runtimes generally ignore NAS NPUs today, so it adds little to chat model speed.
It is a good fit. RAG needs two models in memory at once, one to index and one to answer, plus fast access to the documents, and the N5 Max has both the memory and 5 bays + 5 x M.2. Support is documented by the manufacturer.
The manufacturer also provides AI search across your files (documented by the manufacturer), which covers simple document questions without setting up your own stack.
Networking is 2 x 10GbE. That is ample for indexing large libraries and for several people querying at once.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| 64GB LPDDR5X + 128GB OS storage | 64GB | 128GB | |
| 128GB LPDDR5X + 128GB OS storage | 128GB | 128GB |
The closest Minisforum model by bay count and AINAS Score. Only the specifications that differ are listed.
The N5 Max scores best for Home Lab (72/100) and RAG (70/100). It suits home lab owners who want one box for storage, containers, media and a few AI services.
It is a weaker fit for buyers who expect to add a graphics card later.
Other Native AI NAS systems closest in price.
Buying guides that list it
Workloads it suits
Yes at Q4, but not at Q8. Llama 3.3 70B needs 43GB at Q4 and 75GB at Q8. The Q4 fit is Estimated, a calculation and not a benchmark. Q8 does not fit (Estimated). Qwen2.5 72B is marginal at Q4 (Estimated), meaning it loads with little headroom. Llama 4 Scout 17Bx16E does not fit at Q4 (Estimated).
No. The memory is LPDDR5X 8533 on a 256-bit bus and it is soldered, so it cannot be expanded. The N5 Max is sold with 64GB or 128GB. Choose the size you need at purchase, because it cannot be changed afterwards.
No. There is no PCIe slot and no OCuLink port. The only fast external connections are USB4 ports, one at 40Gbps and two USB4 v2 at 80Gbps. Using an external GPU over them is unconfirmed and not documented by Minisforum, so do not buy on that assumption.
Yes, 5. They are one M.2 2280 PCIe 4.0 x4, one M.2 2230/2280 PCIe 4.0 x1 and three M.2 2280 PCIe 4.0 x1. The OS SSD occupies one of them. Alongside these there are 5 SATA drive bays.
Yes, 50 TOPS. The AMD Ryzen AI Max+ 395 includes a 50 TOPS NPU alongside the Radeon 8060S, and Minisforum markets up to 126 TOPS for the whole platform. Hardware acceleration is vendor confirmed. Minisforum's own local AI stack uses llama.cpp as its core runtime, with Ollama as an optional runtime, and adds the MinisOpenClaw local AI agent and a built-in knowledge base with RAG and citations. All of it runs on the 64GB or 128GB of unified LPDDR5X memory.
Two. One has 64GB LPDDR5X with 128GB of OS storage. The other has 128GB LPDDR5X with 128GB of OS storage. Both use soldered memory, so you cannot change the memory later. The warranty is 2 years.
Yes. Ollama support is documented by the manufacturer. See the capability table above for the source.
Partly. Open WebUI support is not yet verified. See the capability table above for the source.
Yes. AI photos support is documented by the manufacturer. See the capability table above for the source.
No. Surveillance AI support is not yet verified. See the capability table above for the source.
Yes. Hardware acceleration support is documented by the manufacturer. See the capability table above for the source.
It has 2 x 10GbE built in. The fastest port is 2 x 10GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.