M5Stack LLM630 Compute Kit AX630C for Edge AI

The LLM630 Compute Kit is a powerful AI development kit for modern edge computing applications. It is based on the AX630C SoC with integrated NPU (3.2 TOPs@INT8) and is ideal for image processing, speech interaction, and LLM inference directly at the network edge. With 4 GB LPDDR4 RAM (2 GB for hardware acceleration) and 32 GB eMMC, it offers sufficient resources for parallel AI models and stable computing performance. The integrated ESP32-C6 interface with Wi-Fi 6 and a Gigabit Ethernet port allow flexible, fast network connectivity – both wirelessly and via LAN. A versatile socket extends the kit with audio input and output, a camera and display interface, as well as Grove I/O for sensors. Two USB-C ports with OTG function allow the system to easily communicate with peripherals. An integrated battery management system with charging and detection chips allows operation away from the socket. The variety of interfaces (DSI, CSI, UART, I2C, microSD) enables numerous application options in areas such as smart home, security, robotics, or agriculture. Thanks to support for the StackFlow framework, AI models like YOLO, Llama3 or Whisper can be integrated directly – via pipeline or targeted invocation. The kit provides a future-proof platform for developers, schools, and research institutions looking to realize local AI functions.

Recommended Accessories:

• M5Stack Camera Module 8MP for LLM630 Kit SC850SL

Features:

• The Axera AX630C processor provides 3.2 TOPs@INT8 for fast AI inference at the edge.
• An integrated ESP32-C6 chip provides Wi-Fi 6 functionality at 2.4 GHz.
• Thanks to DSI and CSI interfaces, displays and cameras can be connected directly.
• An SMA antenna connection significantly improves wireless signal stability.
• The baseboard offers audio interfaces, sensors, microSD, and Grove ports.
• A battery connection with charging control allows mobile applications without external power supply.
• Two USB-C ports allow both data transfer and OTG operation.
• Compatibility with StackFlow facilitates the integration of complex AI pipelines.

Application Examples:

• AI-driven speech and image processing in smart homes
• Autonomous security or analysis devices for surveillance systems
• Mobile sensor platforms for agriculture with AI evaluation
• Use in education and research for training in embedded AI
• Development of multimodal robot or control systems with local processing

Download and Links:

• Quick Start Guide
• Official M5Stack Forum

Technical Specifications:

• Manufacturer: M5Stack
• Manufacturer SKU: K143
• Processor: AX630C@Dual Cortex A53 1.2 GHz, 3.2 TOPs@INT8
• RAM: 4 GB LPDDR4 (2 GB System + 2 GB NPU)
• eMMC: 32 GB, Standard 5.1
• WLAN: ESP32-C6, Wi-Fi 6 (2.4 GHz)
• LAN: 1× Gigabit Ethernet (JL2101B-N040C)
• Audio: MIC & SPK Header (1.25 mm), Class-D Amplifier
• Sensor: 6-Axis BMI270
• Display: DSI (2 Lane, max. 1080p)
• Camera: CSI (4 Lane, max. 4K@30fps)
• USB: 2× USB-C, OTG-capable
• Storage Expansion: microSD
• Interfaces: Grove (I2C, UART), FUNC Header (Wake, I2C, LED etc.)
• Battery Connection: 1.25 mm 3.7 V (Li-Ion/LiPo), Charging & Detection Chips Integrated
• Operating Temperature: 0 °C to 40 °C
• Dimensions: 133.0 × 83.0 × 17.0 mm
• Weight: 60.0 g

Package Contents:

• 1× LLM630 Compute Kit
• 1× SMA Wi-Fi Antenna
• 1× Heatsink
• 4× EVA Foot Pads