Infini-Megrez is an edge multimodal understanding model developed by Wuwen Xinqun, based on the Megrez-3B-Instruct extension. It excels in comprehending and analyzing three types of modal data: images, text, and audio, achieving optimal accuracy in image understanding, language comprehension, and speech recognition. The model is optimized for a synergistic hardware-software collaboration, ensuring that its structural parameters are highly compatible with mainstream hardware, achieving inference speeds up to 300% faster than similar precision models. It is straightforward to use, based on the original LLaMA architecture, allowing developers to deploy the model on various platforms without modifications, minimizing the complexity of secondary development. Additionally, Infini-Megrez provides a complete WebSearch solution, enabling the model to automatically determine when to trigger search calls, switch between searching and dialogue, and deliver enhanced summarization results.