Specifications
MS-NAT5000
NVIDIA Jetson Thor T5000 module, up to 2,070 FP4 TFLOPS
MS-NAT4000
NVIDIA Jetson Thor T4000 module, up to 1,200 FP4 TFLOPS
MS-NANO
Fanless compact system on NVIDIA Jetson Orin Nano
MS-NANO GPU
1024-core NVIDIA Ampere GPU with 32 Tensor Cores
Memory (T5000)
128GB 256-bit LPDDR5X
Memory (T4000)
64GB 256-bit LPDDR5X
Networking
Dual 5GbE LAN
High-Speed Fabric
QSFP28 connectivity supporting four 25GbE links
Storage
PCIe 5.0 x4 NVMe (M.2)
Expansion
M.2 slots for Wi-Fi, 5G and SSD storage
USB
USB 3.2
Power over Ethernet
Optional PoE on both LAN ports (MS-NANO)
Operating Temperature
-20°C to +50°C (MS-NANO fanless design)
Cooling
Fanless passive cooling (MS-NANO); industrial-grade chassis on NAT series
Target Workloads
Multimodal and generative AI inference, real-time perception and autonomous operation
Award
2026 BC Award winner — Compact Edge AI Platform Powered by NVIDIA Jetson Thor
Debut
LEAP 2026, Riyadh (31 Aug – 3 Sep 2026)
Platform Support
NVIDIA JetPack and Holoscan-ready edge AI software stack
Overview
BIOSTAR's MS-NAT5000 and MS-NAT4000 series bring NVIDIA Jetson Thor T5000 and T4000 modules into a compact, industry-grade edge AI platform. The flagship MS-NAT5000 delivers up to 2,070 FP4 TFLOPS with 128GB of 256-bit LPDDR5X memory, while the MS-NAT4000 pairs the T4000 module with 64GB LPDDR5X and 1,200 FP4 TFLOPS — enough headroom for multimodal and generative AI workloads running entirely on-device.
Connectivity is built for sensor-rich physical AI. Dual 5GbE LAN feeds camera and sensor traffic, and a QSFP28 cage carrying four 25GbE links provides a high-bandwidth uplink or inter-node fabric — an unusual amount of network throughput for a platform in this form factor. Storage hangs off PCIe 5.0 x4 NVMe with additional M.2 slots for Wi-Fi, 5G and SSD expansion.
For compact and cost-sensitive deployments BIOSTAR offers the fanless MS-NANO, built on NVIDIA Jetson Orin Nano with a 1024-core Ampere GPU and 32 Tensor Cores. It adds optional PoE on both LAN ports and operates from -20°C to +50°C, making it suitable for vision sensing systems, autonomous mobile robots and humanoid robot perception stacks where space and power are constrained.
Key Benefits
Flagship Jetson Thor performance — up to 2,070 FP4 TFLOPS on T5000 with 128GB LPDDR5X for large multimodal models at the edge. Exceptional network density with dual 5GbE plus QSFP28 carrying four 25GbE links for sensor aggregation and clustering. PCIe 5.0 x4 NVMe storage for high-throughput model and dataset loading. Scaled-down MS-NANO option brings the same platform family down to a fanless Orin Nano footprint with PoE. Award-recognised design — 2026 BC Award winner for Compact Edge AI Platform Powered by NVIDIA Jetson Thor.
Applications
Humanoid and collaborative robots; autonomous mobile robots and vision sensing systems; smart manufacturing and industrial inspection; smart healthcare and medical imaging; real-time multimodal generative AI at the edge; autonomous machine perception for logistics and agriculture.
Request a Quote — BIOSTAR MS-NAT5000 SERIES — JETSON THOR EDGE AI PLATFORM
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →