According to DeepLearning.AI, Baidu has released Ernie-4.5-VL-28B-A3B-Thinking, an open-weights Mixture-of-Experts (MoE) vision-language model that leads many visual reasoning benchmarks while maintaining low operational costs (source: DeepLearning.AI). In addition, Baidu introduced Ernie-5.0, a proprietary, natively multimodal AI model with 2.4 trillion parameters, positioning it among the largest and most advanced AI models to date (source: DeepLearning.AI). These launches signal significant progress for enterprise AI adoption, offering scalable, high-performance solutions for multimodal applications such as smart search, content moderation, and intelligent customer service. Baidu’s open-weights approach for Ernie-4.5-VL-28B-A3B-Thinking also presents new opportunities for AI developers to build cost-effective vision-language systems in both commercial and research contexts.