Specifications
NPU
Mobilint REGULUS AI SoC
AI Performance
10 TOPS (INT8)
CPU
Quad-core Arm Cortex-A53 up to 1.2 GHz
Power Consumption
2.4 W — engineered for 24/7 operation
Memory
4 GB / 8 GB LPDDR4X at 4,266 MT/s
Data Interface
USB 3.1 Gen1
Dimensions
90 x 40 x 16 mm
Weight
88 g
Ingress Protection
IP65 (TBD)
Host Architectures
x86, ARM, RISC-V
Supported OS
Windows 11, Linux 22.04.5
Supported Frameworks
Keras, TensorFlow, TensorFlow Lite, PyTorch, ONNX
SDK
Mobilint SDK qb
LED Indicators
1 (Power LED)
Deployment Model
Plug-in USB accelerator — no chassis swap, no host PCIe slot required
Model Coverage
490+ open and closed-source models validated on the NPU architecture
Price
Quote upon request
Availability
Available
Overview
The MLD-R1 is a USB 3.1 Gen1 stick built on the REGULUS AI SoC that adds 10 TOPS of INT8 inference to any x86, ARM or RISC-V host, drawing just 2.4 W — a figure Mobilint sizes for continuous 24/7 operation. It requires no chassis change, no PCIe slot and no auxiliary power, which makes it a practical route to on-device inference for existing fleets and for prototype validation.
The stick carries 4 GB or 8 GB of LPDDR4X running at 4,266 MT/s, measures 90 x 40 x 16 mm, weighs 88 g and is specified to IP65 ingress protection. Host support covers Windows 11 and Linux 22.04.5, with Keras, TensorFlow, TensorFlow Lite, PyTorch and ONNX models compiled through the SDK qb driver and runtime.
Because the same REGULUS architecture sits under the MLM-1 SoM, REGULUS AI EVK and MLX-R1 mini AI box, work validated on the MLD-R1 carries over to embedded and vehicle-mounted deployments without re-quantisation or model rewriting.
Typical uses published by Mobilint include portable diagnostic devices running offline, forklift pedestrian and hard-hat detection, AGV and AMR onboard vision and generative AI inference inside a sub-10 W budget, and restricted-zone intrusion detection on construction machinery.
Key Benefits
2.4 W for 10 TOPS — battery-compatible edge inference. No host modification — USB 3.1 Gen1 plug-in. Multi-architecture hosts — x86, ARM and RISC-V. IP65 enclosure for industrial and outdoor environments. Same toolchain as Mobilint SoM and box products.
Applications
Portable and offline diagnostic devices; forklift and AGV safety monitoring; construction machinery hazard detection; industrial worker-safety analytics; engineering-validation platforms for REGULUS-based products; OEM appliances needing bolt-on AI.
Request a Quote — MOBILINT REGULUS MLD-R1 AI USB — 10 TOPS STICK ACCELERATOR AT 2.4 W
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →