Why ZNAS?

Dual-mode AI compute: Rockchip RK3588 (6 TOPS built-in) + 3× stacked iCore-1828 M.2 cards (18 TOPS each = 54 TOPS additional) for a total of 60 TOPS.
27B local LLM: Qwen2.5-32B / DeepSeek-V2 27B / Llama 3.1 run entirely on-device, no internet required, ~30 tokens/second.
6-bay hot-swap RAID: 6× 3.5" SATA drive bays, RAID 0/1/5/6/10/JBOD, up to 132 TB raw capacity, hot-swap with predictive disk failure alerts.
One-command LLM install: bundled Ollama + Open WebUI, model marketplace, multi-user isolation, RESTful API compatible with OpenAI / Claude / Gemini.
All-in-one NAS: SMB/NFS/WebDAV/AFP file sharing, Time Machine, media server (Plex/Jellyfin), Docker + K3s app store.
Triple gigabit network: 2× 2.5 GbE for link aggregation + 1× 10 GbE SFP+ for high-speed NAS link, full-wire-speed local transfer.
8K HDMI output: HDMI 2.1 8K@60 display output built-in, directly drive living-room TV as media center + AI conversational console.
Smart cooling: dual turbo fans + magnetic dust filter + automatic temperature curve, < 28 dB at idle under continuous LLM inference.
Three-year warranty: industrial-grade components, MTBF > 50,000 hours, 7×24 continuous operation.

Specifications

Detailed Specifications

SoC	Rockchip RK3588 (4× Cortex-A76 @2.4 GHz + 4× Cortex-A55 @1.8 GHz, 6 TOPS NPU)
AI Compute	RK3588 + 3× iCore-1828 M.2 cards = 60 TOPS, supports 27B FP16 / 70B INT4 LLMs
Memory	16 GB / 32 GB / 64 GB LPDDR5 (up to 32 GB onboard + 32 GB SO-DIMM)
Storage Bays	6× 3.5" SATA HDD hot-swap, RAID 0/1/5/6/10/JBOD, up to 132 TB
M.2 Slots	2× M.2 NVMe 2280 (system + cache), 3× M.2 Key-M for iCore-1828 cards
Network	2× 2.5 GbE + 1× 10 GbE SFP+ optional + Wi-Fi 6 + BT 5.2
USB/Display	2× USB 3.2 Gen 2 (10 Gbps) + 2× USB 2.0 + HDMI 2.1 (8K) + USB-C DP Alt + 3.5 mm audio
LLM Runtime	Pre-installed Ollama + Open WebUI + RKLLM, supports Qwen2.5-32B / Llama 3.1 / DeepSeek-V2 / Mistral
AI Speed	27B Q4 model: 30+ t/s token throughput (multi-card combined)
Cooling	Dual turbo fans + magnetic dust filter + smart temperature curve, < 28 dB at idle
OS	OpenCloudOS 8 + Docker + K3s + Cockpit web admin; supports Ubuntu 22.04 / Debian 12
Power	Built-in 250 W 80+ Gold ATX PSU, 100-240 V AC wide input
Compliance	CCC + 3C + CE / FCC / RoHS, surge 2 kV, ESD 8 kV contact
Use Cases	Home: local ChatGPT replacement + private doc RAG + auto-transcoding. SMB: replace cloud LLM API + NAS, 12-month TCO payback. Retail: POS backup + edge CV analytics.

Applications

Scenarios

Home — Personal AI assistant + private knowledge base + family photo backup. One device replaces ChatGPT Plus + iCloud + NAS.SMB — Edge AI inference box + on-prem file server. Save $50-200/month cloud LLM API fees, typical 12-month ROI.Retail — POS local backup + edge CV people counting + customer behaviour analytics, no cloud dependency.Education — On-prem LLM for K12 labs (GDPR-friendly), teacher Q&A, document Q&A over school libraries.Healthcare — Clinical note drafting + patient education + medical literature RAG, all on-device (HIPAA-compatible).Developer — Self-hosted LLM API + Docker dev environment + Git LFS storage, zero per-token billing.

ZNAS

Why ZNAS?

Detailed Specifications

Scenarios

Product Gallery