64GB max RAM · 16 bays · 1GbE
AINAS Score 48
No verified price

AI ReadySynology
Why AI Ready: Intel Xeon D-1531 with no NPU, two PCIe Gen3 x8 slots (x8 link) and a 64GB memory ceiling, which meets the AI Ready line for Xeon under 128GB. Synology's card list names network and M.2 cards, not GPUs.
Updated
Can the Synology RS3621RPxs run local AI?
The Synology RS3621RPxs can run small local models, with limits. Its Intel Xeon D-1531 and 64GB memory ceiling suit 32B class models comfortably, and 70B class models at 4-bit with little headroom (Estimated), so treat it as a storage-first NAS that can host light AI tools.
There is no way to add a discrete GPU, so all inference runs on the CPU.
Local LLM support is confirmed by community reports.
The RS3621RPxs suits a small business or home lab that wants a 12-bay rackmount NAS with ECC memory and the option to run a few AI tools alongside file storage. It has an Intel Xeon D-1531 with 6 cores, 8GB of DDR4 ECC UDIMM memory that expands to 64GB, and two PCIe Gen3 x8 slots for network or M.2 cards. Docker and virtual machines are both supported, and Synology gives a 5 year warranty. Surveillance Station is supported for up to 75 IP cameras, and Synology Photos face and object recognition is vendor confirmed. It scores 44 on the AINAS Score, which reflects a capable general NAS with CPU only AI. At the 64GB ceiling, Llama 3.3 70B fits at Q4 (Estimated), though not at Q8 (Estimated).
The one thing that rules out the RS3621RPxs for some buyers is the lack of a GPU path. It has no GPU or NPU, and Synology's GPU card list does not include it, so models use the CPU only. That affects buyers who want local LLMs, private document chat or creative AI to be the main job. Those buyers should look for a system with a supported GPU slot or an NPU, a memory ceiling above 64GB, and 10GbE or faster networking built in. This system has four 1GbE ports and reaches faster speeds only through an added 25GbE or 10GbE card. Models above its memory ceiling, such as Llama 4 Scout 17Bx16E, do not fit at Q4 (Estimated).
No verified price
We have not verified a listing for the Synology RS3621RPxs at a retailer we track. The specifications below stand on their own; a price appears here as soon as a listing is confirmed.
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.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 | Community confirmed | CPU-only Ollama via Container Manager on x86 DSM models. 64 GB memory ceiling, no GPU or NPU. Guide covers the same DSM platform, not this exact unit. |
| Ollama | Yes | Community confirmed | Not in Package Center; ollama/ollama container runs in Container Manager on x86 Synology DSM (CPU only). Guide covers the same DSM platform, not this exact unit. |
| Open WebUI | Yes | Community confirmed | Guide documents Open WebUI alongside Ollama on Synology DSM via Container Manager; same DSM platform, not this exact unit. |
| RAG | Limited | Unverified | No Synology local RAG on this model. AI Console bridges to third-party AI providers. Open WebUI document chat via Docker plausible but not verified on this unit. |
| AI search | No | Vendor confirmed | GPU-powered AI Search in Synology Drive is limited to FS3420, FS6420 and 26-series rack models per the AI Console 2.0 and AI Runtime 1.0 betas; this model is not listed. |
| GPU expansion | No | Vendor confirmed | 2x PCIe Gen3 x8 slots (x8 link) exists but this model is not on the Synology GPU card list, which says unlisted models do not support GPU cards. |
| Hardware acceleration | No | Vendor confirmed | Spec lists a hardware encryption engine only; no integrated GPU, NPU or hardware transcoding listed. |
| AI photos | Yes | Vendor confirmed | Synology Photos face and object recognition listed for this model (on-device, CPU). |
| AI video | Limited | Vendor confirmed | No on-NAS deep video analytics (DVA series only). Surveillance Station can use on-camera AI people and vehicle detection from Synology BC500 cameras. |
| Surveillance AI | Yes | Vendor confirmed | Surveillance Station supported, up to 75 IP cameras (1080p H.264, 2100 FPS); 2 free licences, more require purchase. |
Calculated from the RS3621RPxs'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 RS3621RPxs's use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
Local LLM is rated OK. Ollama runs in Container Manager on the CPU only and is community confirmed, local LLM support is limited, and with 64GB fitted, models up to Llama 3.3 70B fit at Q4 (Estimated).
Private RAG is rated OK, with a caveat. Synology has no local RAG on this model, and document chat through Open WebUI in Docker is plausible but unverified on this unit, although Open WebUI itself is community confirmed alongside Ollama.
AI search is rated poor. Vendor documentation confirms GPU-powered AI Search in Synology Drive is limited to a short list of other rack models, this one is not among them, and it has no GPU or NPU to use.
AI photos is rated OK, using the CPU alone. Synology Photos face and object recognition is vendor confirmed for this model and runs on the CPU, as the system has no NPU or integrated GPU to help.
Surveillance is rated OK. Surveillance Station is vendor confirmed for up to 75 IP cameras at 1080p H.264, with 2 free licences and more available by purchase, but there is no on-NAS deep video analytics.
Home lab is rated OK. Docker and virtual machines are both supported, and the 12 bays, 6-core Xeon D-1531, ECC memory expandable to 64GB and two PCIe Gen3 x8 slots leave room to grow.
Business use is rated OK. ECC memory, a 5 year warranty, 12 bays and a 500W power supply suit file serving, though the network is four 1GbE ports unless you add a 10GbE or 25GbE card in a PCIe slot.
Creative AI is rated OK, but only for CPU work. GPU expansion is vendor confirmed as not supported, there is no hardware acceleration, and local LLM support is limited, so the 64GB memory ceiling is the main strength.
The RS3621RPxs ships with 8GB of DDR4 ECC UDIMM and supports up to 64GB with ECC. Upgrading to the maximum lifts the model ceiling from small models up to around 7B parameters to 32B class models comfortably, and 70B class models at 4-bit with little headroom (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, 64GB leaves space for a model of roughly 93B parameters at most (Estimated).
ECC memory support is a plus for always-on systems that also hold important files.
No. The RS3621RPxs has no PCIe slot or OCuLink port for a graphics card. Inference runs on the CPU, so keep to small models.
With limits. The RS3621RPxs 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.
Networking is 4 x 1GbE. Fine for one or two users; indexing a large archive over the network will be slower.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| RS3621RPxs 8GB ECC | 8GB | None |
The closest Synology model by bay count and AINAS Score. Only the specifications that differ are listed.
The RS3621RPxs scores best for Home Lab (49/100) and Business (47/100). 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 and for teams moving large files over the network every day.
Other AI Ready systems closest in overall score.
Yes, at Q4 only (Estimated). Llama 3.3 70B needs 43GB at Q4, which is within the 64GB ceiling but not the 8GB installed, so the memory must be expanded (Estimated). At Q8 it needs 75GB and does not fit (Estimated). There is no GPU path, so models use the CPU only, through Ollama in Container Manager, which is community confirmed.
Yes. The RS3621RPxs comes with 8GB of DDR4 ECC UDIMM memory and takes up to 64GB, so the memory sits in slots rather than being soldered. The ceiling matters for local models, since the amount of memory decides which models load. Llama 3.3 70B at Q4 is the largest model listed on this page that fits (Estimated), and it needs expanded memory.
No. It has two PCIe Gen3 x8 slots, but the model is not on Synology's GPU card list, which says unlisted models do not support GPU cards. This is vendor confirmed. The cards Synology lists for these slots are a 25GbE card, 10GbE cards and M.2 adapters, so the slots suit networking and M.2 storage, not graphics.
Two cards. The RS3621RPxs has 2x PCIe Gen3 x8 slots with x8 links, and Synology names network and M.2 cards for them: the E25G30-F2 for 25GbE, the E10G30-F2 and other 10GbE cards, and the E10M20-T1 and M2D20 for M.2 storage. Graphics cards are not among them, because the model is not on Synology's GPU card list. With only four 1GbE ports built in, one slot often goes to a faster network card, leaving the other for M.2.
Yes. The RS3621RPxs uses DDR4 ECC UDIMM memory, and the configuration listed is 8GB ECC. Expansion up to 64GB keeps to the same ECC UDIMM type. ECC suits file serving and virtual machines left running for long periods, and Synology confirms Container Manager and Virtual Machine Manager support for this model. It does not add AI capability, which is CPU only.
Yes. Ollama support is confirmed by community reports. See the capability table above for the source.
Yes. Open WebUI support is confirmed by community reports. 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.
No. Hardware acceleration support is documented by the manufacturer. See the capability table above for the source.
It has 4 x 1GbE built in, and the 2x PCIe Gen3 x8 slots (x8 link); Synology-listed cards: E25G30-F2 (25GbE), E10G30-F2 and other 10GbE NICs, E10M20-T1 and M2D20 (M.2) can take a faster network card. The fastest port is 1GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.