Yesterday, the LongCat team from Meituan introduced its latest open-source large-scale model, LongCat-Flash-Omni. Featuring a staggering 560 billion parameters, this model marks significant advancements in comprehensive multimodal capabilities, positioning it as the most sophisticated model in the open-source landscape. As the first open-source large language model in the industry to combine comprehensive multimodal support, an end-to-end framework, and efficient inference with a vast number of parameters, LongCat-Flash-Omni effectively bridges the gap between having a large number of parameters and achieving high efficiency. Its groundbreaking architecture enables low-latency, high-quality audio and video processing, along with seamless streaming speech generation. Additionally, it incorporates a sizable context window and supports prolonged audio and video interactions, showcasing outstanding multimodal prowess. Presently, the official LongCat App is accessible, offering internet search and voice call capabilities.