South Korean company Mobilint, which specializes in neural processor (NPU) development, has unveiled a compact USB artificial intelligence accelerator dubbed the MLD-R1. The device targets edge AI computing, enabling local execution of machine learning models without relying on cloud infrastructure.
10 TOPS of Performance in a Compact Form Factor
At the heart of the MLD-R1 lies Mobilint’s proprietary REGULUS AI SoC, delivering 10 TOPS (INT8) of AI performance. This level is sufficient for a wide range of edge AI workloads, including:
- computer vision;
- object recognition;
- video analytics;
- industrial monitoring;
- smart IoT devices.
Besides the neural engine, the chip integrates a quad-core Arm Cortex-A53 CPU clocked at 1.5 GHz.
Key Specifications of the Mobilint MLD-R1
- AI performance: 10 TOPS (INT8);
- SoC: REGULUS AI;
- CPU: 4 × Arm Cortex-A53 (1.5 GHz);
- Memory: 4 GB or 8 GB LPDDR4X‑4267;
- Interface: USB 3.0 (5 Gb/s);
- Dimensions: 90 × 40 × 16 mm;
- TDP: only 3 W;
- Protection: IP65 (dust and water resistant);
- Operating mode: 24/7 continuous operation.
Its low power consumption makes the MLD‑R1 suitable for round‑the‑clock deployment in industrial, transportation, and urban infrastructure systems.
Broad AI Framework Support
Mobilint claims compatibility with most major machine‑learning platforms:
- TensorFlow;
- TensorFlow Lite;
- PyTorch;
- ONNX;
- Keras.
The MLD‑R1 runs under both Windows and Linux and supports processor architectures including:
- x86;
- Arm;
- RISC‑V.
Why a USB AI Accelerator Matters
Devices like the MLD‑R1 shift AI inference directly to a local PC or embedded system, providing:
- minimal processing latency;
- operation without internet access;
- enhanced data confidentiality;
- reduced load on the host CPU.
Such solutions are increasingly adopted in manufacturing, video surveillance, robotics, medical equipment, and edge AI scenarios.
HiTech Expert Analysis
The edge AI market is rapidly moving away from bulky GPUs toward compact, specialized NPU accelerators. With the rise of agent‑based AI systems, more models are executed directly on devices, reducing dependence on cloud services and cutting inference costs.
Although 10 TOPS cannot be matched against the performance of modern discrete GPUs or high‑end AI accelerators, it is ample for most typical edge AI applications such as video analytics, factory automation, robotics, and IoT. Combined with a mere 3 W TDP, a compact footprint, and an IP65 rating, the Mobilint MLD‑R1 presents an attractive option for industrial systems that require local AI execution without the need for constant connectivity to data centers (see related article).
Mobilint unveiled the personal USB AI accelerator MLD‑R1 – read the full story on HiTech.Expert.
