On March 31, during Baidu AI DAY, Baidu introduced the industry's pioneering end-to-end speech-language large model, leveraging the innovative Cross-Attention technology. This groundbreaking model boasts ultra-low latency and cost efficiency, particularly in telephone-based speech Q&A scenarios, where it slashes invocation costs by 50% to 90% compared to industry standards. On the same occasion, the Wenxiaoyan brand announced its revitalization and pioneering integration of this large model, ushering in advancements such as multi-model fusion scheduling and image Q&A capabilities. The revamped Wenxiaoyan now offers more lifelike voice chat effects and support for multiple dialects, markedly enhancing response speed and task handling proficiency.