Description The LLM630 Compute Kit is an edge AI inference development platform for large language model and computer vision applications. Its mainboard uses an Axera AX630C SoC with a high-efficiency NPU that delivers 3.2 TOPs at INT8 precision, supporting local inference for CV, LLM, speech recognition, text-to-speech, and wake word workflows. The LLM630 Compute Kit combines 4GB LPDDR4 memory, 32GB eMMC storage, JL2101B-N040C gigabit Ethernet, and an ESP32-C6 2.4GHz Wi-Fi 6 network card with Wi-Fi and Ethernet bridging. The baseboard adds BMI270 6-axis sensing, microphone and speaker headers, MIPI DSI display, MIPI CSI camera, microSD, USB OTG, charging and battery detection, and two HY2.0-4P expansion ports. With the StackFlow framework, the LLM630 Compute Kit can run AI tasks through separate invocation or pipeline flow. Documented model examples include Yolov11, DepthAnything, InternVL2.5-1B, Qwen2.5-0.5B/1.5B, Llama3.2-1B, Whisper, and MeloTTS for interactive edge intelligence. Features LLM630 Compute Kit uses the Axera AX630C SoC for local edge AI inference. 3.2 TOPs INT8 NPU performance, with up to 12.8 TOPs at INT4 precision. 4GB LPDDR4 memory and 32GB eMMC support model storage and parallel loading. JL2101B-N040C gigabit Ethernet plus ESP32-C6 Wi-Fi 6 at 2.4GHz. USB OTG, two USB Type-C ports, microSD expansion, and SMA antenna interface. MIPI DSI display interface and MIPI CSI camera interface for vision applications. Microphone and speaker headers with NS4150B Class D amplifier for voice interaction. Battery connector, AW32001E charging chip, and BQ27220YZFR battery detection chip. StackFlow development framework for CV, LLM, speech, and multimodal AI pipelines. Specifications SoC AX630C, dual-core Cortex-A53, 1.2GHz main frequency NPU 3.2 TOPs @ INT8, max 12.8 TOPs @ INT4 RAM 4GB LPDDR4, with 2GB system memory and 2GB hardware acceleration memory Storage eMMC 5.1, 32GB Wired network JL2101B-N040C, 1GbE Wireless network ESP32-C6, 2.4GHz Wi-Fi USB-UART CH9102F USB to serial port USB-OTG USB 2.0 host or device Antenna interface SMA inner hole Audio interface MIC and SPK header, 5P @ 1.25mm Display interface MIPI DSI, 1 x 2-lane, max 1080p @ 30fps, 1.25mm Camera interface MIPI CSI, 1 x 4-lane, max 4K @ 30fps, 1.25mm Connector model Audio interface 1.25-5P; function expansion interface 1.25-8P; battery power interface 1.25-2P Battery management AW32001ECSR charging chip, BQ27220YZFR power detection chip Battery connector 1.25mm battery connector for 3.7V lithium battery USB ports 2 x USB Type-C ports for data transfer and OTG function USB power input 5V@2A USB Type-C header PortA 4P @ 2.0mm for I2C; PortC 4P @ 2.0mm for UART External storage microSD card slot External function interface FUNC header 8P @ 1.25mm for wake-up, power management, external LED control, and I2C Buttons 2 buttons for power on/off, user interaction, and reset functions Sensor BMI270 6-axis sensor Operating temperature 0 C to 40 C Product size 90.3 x 31.6 x 12.0mm Package size 133.0 x 83.0 x 17.0mm Product weight 23.6g Gross weight 68.1g Applications Security monitoring with local vision model inference. Smart home control panels and edge AI gateways. Smart agriculture nodes that combine camera, sensor, and network data. Smart sales terminals with speech and visual interaction. Interactive robots using local LLM, voice, and camera pipelines. Education and learning platforms built around the LLM630 Compute Kit. Product List 1 x LLM630 Compute Kit 1 x SMA 2.4GHz Wi-Fi antenna 1 x heatsink 4 x EVA foam foot pads Other Content Power behavior depends on the power source. When connected to USB power, long press the PWR/BOOT button until the blue light flashes to enter download mode, and short press the PWR/BOOT button to reset. When only a battery is connected, short press PWR/BOOT to power on or reset, and double-click PWR/BOOT to power off. The LLM630 Compute Kit is designed for expandable edge intelligence. The DSI and CSI connectors support display and camera development, the audio headers support voice input and output, and the FUNC header provides wake-up, power-management, external LED, and I2C functions for custom carrier designs. Resources LLM630 Compute Kit Documentation LLM630 Compute Kit Schematic PDF LLM630 Computer Kit STEP Model File BMI270 6-Axis Sensor Datasheet PDF AW32001E Battery Charger Datasheet PDF NS4150B Class D Amplifier Datasheet PDF BQ27220YZFR Battery Monitor Datasheet PDF AX630C Processor Datasheet PDF ADB, SSH, and UART Connection Debugging Guide Software Package Application Update Guide Image Base Package Update Guide StackFlow API LLM Demo Guide Jupyter Notebook Guide StackFlow AI Overview StackFlow Framework GitHub Repository StackFlow JSON API Documentation FAQ What AI tasks can it run? It can be used for local computer vision, large language model, speech recognition, text-to-speech, wake word, and multimodal pipeline applications through StackFlow. Does it support wired and wireless networking? Yes. The board includes gigabit Ethernet and ESP32-C6 2.4GHz Wi-Fi 6, with Wi-Fi and Ethernet bridging functionality. What peripherals can be expanded? The board exposes MIPI DSI, MIPI CSI, audio headers, two HY2.0-4P ports, microSD, USB OTG, and the FUNC expansion header. What is included in the package? The package includes the LLM630 Compute Kit, SMA Wi-Fi antenna, heatsink, and four EVA foam foot pads.
