K144 M5Stack LLM (large language model) module kit (AX630C)

Module LLM Kit is an intelligent modular kit that focuses on offline AI inference and data communication interface applications. It integrates Module LLM and Module13.2 LLM Mate modules to fulfill offline AI inference and data interaction requirements in various scenarios.

Module LLM is an integrated offline LLM (Large Language Model) inference module specifically designed for end devices that require efficient and intelligent interaction. Whether for smart home applications, voice assistants or industrial controls, Module LLM provides a smooth and natural AI experience without relying on the cloud, ensuring privacy, security and stability.

Module13.2 LLM Mate module provides a variety of interface functions to facilitate system integration and expansion. It achieves stacked power supply with the LLM module via the M5BUS interface; its built-in CH340N USB conversion chip provides USB-to-serial debugging functionality, while the Type-C interface is used for USB log output. In addition, the RJ45 interface works with the integrated network transformer to extend a 100 Mbps Ethernet port and a serial core port (to support SBC applications); the FPC-8P interface connects directly to the LLM module and ensures stable serial communication; and an HT3.96*9P solder pad is reserved for DIY expansion.

The LLM module integrates the StackFlow framework together with the Arduino/UiFlow libraries so that Edge Intelligence can be implemented with just a few lines of code. Powered by the advanced AiXin AX630C SoC processor and with a highly efficient NPU delivering 3.2 TOPS with native support for Transformer models, it effortlessly handles complex AI tasks. Equipped with 4 GB LPDDR4 memory (1 GB for user applications and 3 GB for hardware acceleration) and 32 GB eMMC memory, it supports parallel loading of multiple models and chained inferences, ensuring smooth multitasking. With an operating power consumption of only around 1.5 W, it is far more energy efficient than comparable products and therefore ideal for long-term operation.

The LLM module is compatible with multiple models and comes pre-installed with the Qwen2.5-0.5B large speech model, which has built-in features such as KWS (wake word), ASR (speech recognition), LLM (large speech model) and TTS (text-to-speech). It also supports apt-based fast updates of software and model packages. By installing the openai-api plugin, it becomes compatible with the OpenAI standard API and supports chat, conversation completion, speech-to-text and text-to-speech in addition to various application modes. The official apt repository offers a wealth of large model resources - including deepseek-r1-distill-qwen-1.5b, InternVL25-1B-MPO, Llama-3.2-1B, Qwen2.5-0.5B and Qwen2.5-1.5B - as well as text-to-speech models (whisper-tiny, whisper-base, melotts) and visual models (such as yolo11 and other SOTA models). The repository is constantly updated to support the most modern model applications and to meet the requirements of complex AI tasks.

The LLM module kit is plug-and-play and provides an instant interactive AI experience in conjunction with the M5 host. Users can seamlessly integrate it into existing smart devices without cumbersome setup, quickly enabling smart features and improving device performance. This product is ideal for offline voice assistants, text-to-speech conversion, smart home control, interactive robots and more.

Key features