128GB max RAM · 16 bays · 2 x 10GbE
AINAS Score 68
No verified price

AI ReadyMinisforum
Why AI Ready: AMD Ryzen 5 7640HS (6C/12T) has a 10 TOPS NPU and Radeon 760M, with OCuLink plus PCIe x4 slot and up to 96GB DDR5, but Minisforum does not position local LLMs on it, so ready rather than native.
Updated
Can the Minisforum N5 Air-7640 run local AI?
Only lightly. Expect 70B class models at 4-bit with room left for longer context (Estimated) on the Minisforum N5 Air-7640: AMD Ryzen 5 7640HS does the work with up to 96GB of memory, which is fine for experiments rather than a daily assistant.
Its OCuLink (external GPU) leaves room to add a GPU when you outgrow CPU inference.
Local LLM support is not yet verified.
The N5 Air-7640 suits buyers who want a five-bay home or small-office NAS with a route into AI work later. It has an AMD Ryzen 5 7640HS with 6 cores, a Radeon 760M and a 10 TOPS NPU. Memory is two DDR5 5600 SO-DIMM slots, expandable to 96GB, and it ships without any RAM. Storage is five SATA bays plus 3 M.2 slots. Networking is one 10GbE and one 5GbE port. For graphics expansion it has an OCuLink port and a PCIe x16 slot running at PCIe 4.0 x4, and the vendor confirms both. It runs MinisCloud OS with Docker, and the vendor confirms AI Photo Library with face recognition and semantic search. It has an AINAS Score of 69.
The main catch is that local LLM use is unconfirmed. Minisforum does not document local LLMs, Ollama, Open WebUI or RAG on this model, and all four are Unverified. The memory would hold large models on paper (Estimated), but that is a calculation, not a tested result. This affects anyone buying mainly to run a private assistant or document search. Those buyers should look for a system where the vendor documents local LLM support, container support, and a stated GPU or NPU path in the operating system. The 10 TOPS NPU is present but undocumented in MinisCloud OS, so it adds no confirmed AI capability.
Amazon US
Barebone + 64GB OS storage
Ships from the US; delivery and import charges shown at checkout
$495.00
Verified
In stock
Buy on AmazonAmazon US
16GB DDR5 4800 + 64GB OS storage
Ships from the US; delivery and import charges shown at checkout
$679.00
Verified
In 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 | Limited | Unverified | 10 TOPS NPU and up to 96GB DDR5, but Minisforum does not document local LLMs on this model. |
| Ollama | Limited | Unverified | Docker supported in MinisCloud OS; no Ollama source for this model. |
| Open WebUI | Limited | Unverified | Not documented. |
| RAG | Limited | Unverified | Not documented. |
| AI search | Yes | Vendor confirmed | AI Photo Library semantic image search (MinisCloud OS). |
| GPU expansion | Yes | Vendor confirmed | OCuLink port and internal PCIe x16 slot (PCIe 4.0 x4). |
| Hardware acceleration | Limited | Vendor confirmed | Radeon 760M iGPU and 10 TOPS NPU (AMD); NPU use in MinisCloud OS not documented. |
| AI photos | Yes | Vendor confirmed | AI Photo Library: face recognition, scene sorting, image/semantic search. |
| AI video | No | Unverified | No video AI documented. |
| Surveillance AI | No | Unverified | No surveillance app documented for MinisCloud OS. |
Calculated from the N5 Air-7640'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 | Q5 | Yes | 68GB | Not measured | Estimated |
| Llama 4 Scout 17Bx16ELlama, 109B | Q4 | Yes | 89GB | Not measured | Estimated |
| Qwen2.5 72BQwen, 72.7B | Q5 | Yes | 73GB | Not measured | Estimated |
| Mistral Large 2 123BMistral, 123B | Q4 | Yes | 96GB | Not measured | Estimated |
One verdict per workload, from the N5 Air-7640's use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
We cannot confirm the N5 Air-7640 for local LLM work. The local LLM row is limited and Unverified, with a 10 TOPS NPU and up to 96GB DDR5 but no vendor documentation of local LLMs on this model, so any fit is only an Estimated calculation.
Private RAG is unconfirmed on the N5 Air-7640. The RAG and Open WebUI rows are limited and Unverified, and the notes say not documented, while Docker is supported in MinisCloud OS but the Ollama row has no source for this model.
AI search is a good fit on the N5 Air-7640. The AI search row is Vendor confirmed and covers semantic image search in the AI Photo Library in MinisCloud OS, so it applies to your photo collection rather than general document search.
AI photos is a good fit. The AI photos row is Vendor confirmed, with face recognition, scene sorting and image search in the AI Photo Library, and the five SATA bays plus 3 M.2 slots give room for a large library.
Surveillance is a poor fit for the N5 Air-7640. The surveillance row is marked no and Unverified, and the notes say no surveillance app is documented for MinisCloud OS, so there is no confirmed camera software.
Home lab use is a good fit. Docker is supported, there are 3 M.2 slots, 10GbE and 5GbE ports and two USB4 ports, and five SATA bays hold bulk storage, though virtual machine support is not stated.
Business use is a good fit. Five SATA bays, a 280W power supply, a 2-year warranty and 10GbE and 5GbE networking suit a small office file server, though the memory is non-ECC and the barebone version ships without RAM.
Creative AI is a good fit. The GPU expansion row is Vendor confirmed, covering an OCuLink port and a PCIe x16 slot at PCIe 4.0 x4, while hardware acceleration is limited because NPU use in MinisCloud OS is not documented.
The N5 Air-7640 can be bought without memory and takes up to 96GB of DDR5 5600 SO-DIMM. Fitting the maximum is what turns it into a serious local AI machine.
Our estimate for the biggest model that fits in 96GB is about 146B parameters at 4-bit, after setting aside 8GB for the system. Long context windows reduce that figure (Estimated).
Yes, externally. The N5 Air-7640 has an OCuLink port for an external GPU dock with a direct PCIe link. The dock and its power supply sit outside the case.
The 10 TOPS NPU helps with photo recognition, transcription and vendor AI features. Most open-source runtimes such as Ollama do not use NAS NPUs yet, so do not count it towards language model speed.
As the storage half of a RAG setup it works well; as the model host it is limited. Small embedding models run fine on the N5 Air-7640, while the answering model needs more compute than it has.
The manufacturer also provides AI search across your files (documented by the manufacturer), which covers simple document questions without setting up your own stack.
For network access it has 10GbE + 5GbE. Fast enough that the network will not be the bottleneck.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| Barebone + 64GB OS storage | None | 64GB | |
| 16GB DDR5 4800 + 64GB OS storage | 16GB | 64GB |
The closest Minisforum model by bay count and AINAS Score. Only the specifications that differ are listed.
Best fit: small offices that need shared storage with private AI tools and room to grow. That is where the N5 Air-7640 scores highest (Business 72/100), followed by Local LLM (71/100).
Other AI Ready systems closest in price.
Buying guides that list it
Workloads it suits
Yes, on paper. Mistral Large 2 123B needs 73GB at Q4 and fits at that level (Estimated), against a 96GB memory maximum. It does not fit at Q8, which needs 130GB (Estimated). Qwen3 235B A22B does not fit at Q4 (Estimated). These are calculations, not benchmarks, and the local LLM row is limited and Unverified, so Minisforum has not confirmed local LLM use on this model.
Up to 96GB. It uses DDR5 5600 SO-DIMM across two slots rather than soldered memory, so you can add or swap modules. The base version comes barebone with no RAM, and a 16GB DDR5 4800 configuration is also listed. The memory is non-ECC, so there is no error correction.
Yes, by two routes. It has an OCuLink port and an internal PCIe x16 slot that runs at PCIe 4.0 x4 electrical. The vendor confirms GPU expansion. The page does not list tested cards, so check a card's fit and power needs yourself. The system has a 280W power supply.
Yes, three. Slot A runs at PCIe 4.0 x1, slot B at x2 and slot C at x4, and the 64GB OS SSD occupies the first slot, so two are free for your own drives. They sit alongside the five SATA bays. Both configurations include that 64GB of OS storage, one as a barebone and one with 16GB of DDR5 4800 memory. The x4 slot is the fastest place for model files or an app drive.
Yes. Docker is supported in MinisCloud OS, and Minisforum markets Docker apps on its global page. Virtual machine support is not stated, so do not assume it. Windows 11 Pro and Linux are also listed as operating system options by the manufacturer, if you prefer to install your own.
Yes. The Ryzen 5 7640HS includes a 10 TOPS NPU, alongside a Radeon 760M integrated GPU. The hardware acceleration row is limited, because NPU use in MinisCloud OS is not documented. Treat the NPU as present but not confirmed as usable for AI tasks in the bundled software.
Partly. Ollama support is not yet verified. 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.
Partly. Hardware acceleration support is documented by the manufacturer. See the capability table above for the source.
It has 10GbE + 5GbE built in, and the 1x PCIe x16 slot (PCIe 4.0 x4 electrical) can take a faster network card. The fastest port is 10GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.