Meta 发布轻量级 Muse Ember 模型，用于代理执行
meta 发布了 muse ember，一种用于代理执行的全新轻量级人工智能模型，旨在实现复杂推理，成本效率高，每百万输入令牌仅需 10 美分。
meta's superintelligence lab released muse ember, their lightest and most cost-efficient ai model yet, designed for complex reasoning execution and repetitive tasks in agents. ember costs 10 cents per million input tokens, making it cheaper than frontier competitors like haiku and luna, and competitive with open-source models like deepseek v4.1 flash and glm 5.3 flash. it is trained to work with various media types, including text, images, and audio, and is meant to handle high-volume use cases when paired with a larger model like muse wonder.