Products

Developers

Resources

Company

Products

Developers

Resources

Company

Products

Developers

Resources

Company

MLM-1 SoM

MLM-1 SoM

MLM-1 SoM

REGULUS-Powered System-on-Module for Embedded AI

REGULUS-Powered System-on-Module for Embedded AI

REGULUS-Powered System-on-Module for Embedded AI

Complete AI Compute on a Compact Module

Complete AI Compute on a Compact Module

Complete AI Compute on a Compact Module

MLM-1 SoM is a REGULUS-powered system-on-module that integrates NPU, CPU, ISP, video codec, memory and storage on a single board, powering 10 TOPS of AI inference.

MLM-1 SoM is a REGULUS-powered system-on-module that integrates NPU, CPU, ISP, video codec, memory and storage on a single board, powering 10 TOPS of AI inference.

At 55 × 50 mm with a 5V input, it provides a complete AI compute platform for product teams designing AI into size- and power-constrained devices.

At 55 × 50 mm with a 5V input, it provides a complete AI compute platform for product teams designing AI into size- and power-constrained devices.

Specifications

Specifications

Specifications

MLM-1 SoM

NPU SoC

REGULUS

CPU

Quad-core ARM Cortex-A53 + Cortex-M0+

AI performance

10 TOPS (Dense, INT8)

TDP

3 W (REGULUS), 5 W (SoM)

Memory

8 GB LPDDR4X

Storage

32 GB eMMC 5.1, 256 Mbit SPI flash, microSD

Multimedia

  • Video

    • MJPEG Encoder/Decoder (4K @ 120 FPS)

    • H.265/4 Encoder/Decoder (4K @ 60 FPS)

  • Image

    • Supports ISP (Up to 8M @ 60 FPS)

Supported Connectivity

  • Gigabit Ethernet

  • USB 3.1

  • MIPI CSI-2

  • MIPI DSI

  • SDIO

  • GPIO, I2C, SPI, I2S, UART

  • PWM/PIT

  • JTAG

Connectors

3× 100-pin (J1, J2, J3)

Power

5V input to SoM (DC +12V for evaluation board)

Dimensions

55×50×8 mm (without heatsink)

Weight

  • 22.0 g (without heatsink)

  • 29.7 g (with heatsink)

More frames per watt, where watts are the limit.

More frames per watt, where watts are the limit.

More frames per watt, where watts are the limit.

Up to 4× less power per frame compared to GPU.

REGULUS

GPU A

FPS per watt (higher is better)

MobileNetV2

MobileNetV2

4.2× higher FPS/W

246.6

58.1

YOLOv12n

YOLOv12n

3.3× higher FPS/W

14.7

4.5

YOLOv12l

YOLOv12l

2.8× higher FPS/W

2.4

0.9

Mobilint internal benchmarks, single batch.

Prototype MLM-1 SoM on the Official Evaluation Board.

Prototype MLM-1 SoM on the Official Evaluation Board.

Prototype MLM-1 SoM on the Official Evaluation Board.

REGULUS AI EVK, the official evaluation kit for MLM-1 SoM, is designed as a development bridge for validating your application on the module.

REGULUS AI EVK, the official evaluation kit for MLM-1 SoM, is designed as a development bridge for validating your application on the module.

Develop and validate on the EVK’s carrier board, then design a custom carrier for the target product. The compute module, software stack and compiled models carry over with no changes.

Develop and validate on the EVK’s carrier board, then design a custom carrier for the target product. The compute module, software stack and compiled models carry over with no changes.

Learn more about the REGULUS AI EVK

Learn more about the REGULUS AI EVK