Meta unveils Muse Spark

Meta just put a new brain into its apps at the same time it started moving engineers into a new artificial intelligence unit by memo, not by volunteer sign-up. The model is called Muse Spark, and Meta said on April 8 that it already powers the Meta AI app and website. (about.fb.com) A large language model is the text engine behind a chatbot, the part that predicts the next word the way your phone predicts the next word in a message, but at far bigger scale. Meta says Muse Spark is its most powerful model yet and the first model from Meta Superintelligence Labs. (about.fb.com) Meta is not keeping Muse Spark inside one app. The company said the model will roll out in the coming weeks to WhatsApp, Instagram, Facebook, Messenger, and its artificial intelligence glasses, which means one model is being used as shared plumbing across nearly every big Meta product. (about.fb.com) This is a break from the Meta story people knew last year. Meta built its recent reputation in artificial intelligence around Llama, a family of models it often released with open weights, while Muse Spark is being offered only through Meta products and a private preview application programming interface for selected partners. (about.fb.com) (artificialanalysis.ai) Meta says Muse Spark was built to prioritize “people,” which in practice means the model is tuned for recommendations, posts, and content shared across Instagram, Facebook, and Threads. That is a very different target from a general lab benchmark race, because Meta makes its money inside consumer apps with billions of users and ads around them. (about.fb.com) (cnbc.com) The timing points to a bigger reset inside the company. Reuters reported on April 9 that Meta has been drafting top software engineers from across the company into a new Applied AI Engineering group led by Vice President Maher Saba after first asking for volunteers in March. (froggyweb.com) (timesofindia.indiatimes.com) That group is not just building demos for users. Reuters said the unit is meant to build artificial intelligence agents that could eventually handle much of Meta’s own engineering work, so the company is trying to use artificial intelligence both as a product feature and as an internal labor tool. (timesofindia.indiatimes.com) There is also a leadership story here. CNBC reported that Muse Spark was spearheaded by Alexandr Wang, the Scale AI co-founder Meta brought in to lead Meta Superintelligence Labs, and Axios reported the model was code-named Avocado and built over about nine months. (cnbc.com) (axios.com) So Muse Spark is not just another model launch. It is Meta saying the next phase is closed-weight models inside its own apps, a new lab under new leadership, and a forced shift of engineering talent toward tools that can automate more of the company’s own work. (about.fb.com) (froggyweb.com)