64GB max RAM · 12 x M.2 · 2 x 10GbE
AINAS Score 39
£1,339
Amazon UK · Verified

AI CapableQNAP
Why AI Capable: Atom C5125 is not a current desktop-class CPU; its two PCIe Gen3 x4 slots take only low-profile GPUs up to 6GB on QNAP's list, below the 8GB needed for useful local models.
Updated
Can the QNAP TS-855X run local AI?
Yes, the QNAP TS-855X can run local language models. It pairs Intel Atom C5125 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).
It also has a GPU path (PCIe x4 slot), so a graphics card can be added later for much faster inference.
Local LLM support is documented by the manufacturer.
The TS-855X suits a buyer who wants an eight-bay QNAP tower that can also run local AI containers. It has 8 bays (6 SATA 3.5in and 2 SATA 2.5in SSD), 2 M.2 slots, 8GB of DDR4 UDIMM that goes up to 128GB, and ECC support. Networking is 1 x 10GbE and 2 x 2.5GbE. It runs Docker and virtual machines, and QNAP confirms Ollama and Open WebUI. Two PCIe Gen3 x4 slots take a GPU from QNAP's list. With the full 128GB, Llama 3.3 70B fits at Q4 (Estimated), and so does Mistral Large 2 123B. The AINAS Score is 58.
The limit is the GPU path. The Atom C5125 has no integrated GPU and no NPU, and the supported cards top out at 6GB of VRAM. QNAP's own guide says small models run on the GPU and larger ones run slowly on the CPU. Buyers who want fast large-model inference should look for a system with a GPU path that takes cards with much more VRAM, or a much stronger CPU. Card size is a second limit: slot 1 takes a card up to 260 x 111.15mm, while slot 2 is limited to 150 x 68.9mm.
Amazon US
TS-855X-8G
Ships from the US; delivery and import charges shown at checkout
$1,745.99
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 | Yes | Vendor confirmed | QNAP Ollama guide (x86, optional NVIDIA GPU from compatibility list). Up to 128 GB RAM; supported GPUs top out at 6 GB VRAM, so small models on GPU, larger ones slow on CPU. |
| Ollama | Yes | Vendor confirmed | Container Station Ollama tutorial for x86_64 QTS/QuTS hero, NVIDIA GPU in Container Station mode. |
| Open WebUI | Yes | Vendor confirmed | QNAP Open WebUI tutorial, QTS 5.2 / QuTS hero h5.2+. |
| RAG | Limited | Unverified | Qsirch AI Knowledge Search (RAG) uses cloud LLMs; not confirmed for this model. |
| AI search | Yes | Vendor confirmed | Qsirch AI Semantic Search needs 64-bit x86 and 8 GB RAM. AI Mode needs x74 series, so not available. |
| GPU expansion | Yes | Vendor confirmed | GPGPU list up to RTX 3050 LP 6G and RTX A400 4GB; PCIe Gen3 x4 slots. |
| Hardware acceleration | Limited | Vendor confirmed | No iGPU; Intel QAT for compression/crypto. Coral M.2 and USB Edge TPU on compatibility list. |
| AI photos | Yes | Vendor confirmed | QuMagie AI-assisted photo management (QTS only per spec page); Edge TPU speeds recognition. |
| AI video | Yes | Vendor confirmed | Product page: QVR Elite smart facial recognition for retail and door access. |
| Surveillance AI | Yes | Vendor confirmed | QVR Pro 8 free channels (QTS); QVR Surveillance 2 free + 6 on registration; product page cites QVR Elite 2 free channels. |
Calculated from the TS-855X'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 | Q8 | Yes | 98GB | Not measured | Estimated |
| Llama 4 Scout 17Bx16ELlama, 109B | Q4 | Yes | 89GB | Not measured | Estimated |
| Qwen2.5 72BQwen, 72.7B | Q8 | Yes | 101GB | Not measured | Estimated |
| Mistral Large 2 123BMistral, 123B | Q5 | Yes | 112GB | Not measured | Estimated |
One verdict per workload, from the TS-855X's use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
The TS-855X is a good fit for local LLMs. QNAP confirms Ollama through Container Station, and up to 128GB of RAM lets Llama 3.3 70B fit at Q4 (Estimated), though GPUs are limited to 6GB of VRAM.
Private RAG is workable but not a strong match. Qsirch AI Knowledge Search uses cloud LLMs and is unverified for this model, though Open WebUI with Ollama is vendor confirmed.
AI search is a good fit. QNAP confirms Qsirch AI Semantic Search, which needs 64-bit x86 and 8GB of RAM, and the 8GB included meets that; AI Mode needs an x74 series model, so it is not supported here.
AI photo management is a good fit. QuMagie is vendor confirmed, but QNAP lists it for QTS only, and an Edge TPU can speed up recognition.
Surveillance is a good fit. QVR Pro gives 8 free channels on QTS, QVR Surveillance gives 2 free plus 6 on registration, and QVR Elite handles facial recognition.
Home lab use is a good fit. Docker and virtual machines are both supported, and GPU pass-through is listed as working, with 2 M.2 slots and 2 PCIe Gen3 x4 slots for expansion.
Business use is a good fit. It offers ECC memory support, 8 bays, a 10GbE port with two 2.5GbE ports, and a 3 year warranty.
Creative AI is rated a good fit. GPU expansion is vendor confirmed, but cards stop at RTX 3050 LP 6G, and hardware acceleration is limited, with no iGPU.
The TS-855X ships with 8GB of DDR4 UDIMM and supports up to 128GB with ECC. Upgrading to the maximum lifts the model ceiling from small models up to around 7B parameters to 70B class models at higher-quality quantization, or large mixture-of-experts models at 4-bit (both Estimated).
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).
ECC memory support is a plus for always-on systems that also hold important files.
Partly. The TS-855X has a 2x PCIe Gen3 x4, which suits low-profile cards and network or storage upgrades. Only small single-slot GPUs fit, so expect modest gains.
With limits. The TS-855X can host a small RAG stack such as AnythingLLM or Open WebUI with a compact embedding model, but memory and compute restrict how large the answering model can be. It suits a personal document library more than a team knowledge base.
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 10GbE + 2 x 2.5GbE. That is ample for indexing large libraries and for several people querying at once.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| TS-855X-8G | 8GB | None |
The closest QNAP model by bay count and AINAS Score. Only the specifications that differ are listed.
The TS-855X scores best for Home Lab (65/100) and Business (63/100). It suits home lab owners who want one box for storage, containers, media and a few AI services.
Other AI Capable systems closest in price.
Buying guides that list it
Workloads it suits
Yes, in estimate only. Mistral Large 2 123B needs 73GB at Q4, and the TS-855X takes up to 128GB of RAM, so it fits at Q4 (Estimated). It does not fit at Q8 (Estimated). That is a calculation, not a benchmark. With the 8GB included, you would need to add memory. QNAP says larger models run slowly on the CPU.
Yes. It ships with 8GB of DDR4 UDIMM and has 4 slots, taking up to 128GB (4x 32GB). QNAP's compatibility list shows DDR4-3200 UDIMM modules, both ECC and non-ECC. The memory sits in slots rather than being soldered, so you can fit larger memory later to hold bigger models.
Yes. ECC memory is supported, but the 8GB module it ships with is non-ECC. For ECC you need to fit ECC DDR4 UDIMM modules from QNAP's compatibility list. The list covers both ECC and non-ECC modules at DDR4-3200.
Yes. It has two PCIe Gen3 x4 slots, and QNAP confirms GPU expansion. The listed cards include the GT 1030, GTX 1050 Ti, GTX 1650 LP, RTX 3050 LP 6G, Quadro P620 and P1000, and RTX A400 4GB. The largest is 6GB of VRAM, and slot sizes limit card length.
Yes. Container Station runs Docker on QTS 6.1.0 and QuTS hero h6.1.0, and virtual machines are supported too. QNAP's Ollama tutorial is vendor confirmed for x86_64 and uses an NVIDIA GPU in Container Station mode, so a card from the compatibility list can serve small models while larger ones run on the 8-core Atom C5125. Open WebUI is vendor confirmed for QTS 5.2 and QuTS hero h5.2 or later. With 8GB included, memory is the first upgrade for anything beyond small containers.
Yes, 2. They are M.2 2280 PCIe Gen3 x4 slots, separate from the 8 drive bays (6 SATA 3.5in and 2 SATA 2.5in SSD). QNAP also lists a Coral M.2 accelerator as compatible.
Yes. Ollama support is documented by the manufacturer. See the capability table above for the source.
Yes. Open WebUI support is documented by the manufacturer. 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.
Yes. Surveillance AI support is documented by the manufacturer. 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 + 2 x 2.5GbE built in, and the 2x PCIe Gen3 x4 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.