Meta launches its smallest and most efficient model, Muse Ember, aimed at reasoning execution and daily agent work.

Meta unveils its smallest and most efficient model, Muse Ember, designed for complex reasoning tasks and routine agent work. It can execute workflows across applications, call tools in sequence, run tests, and perform targeted code fixes. With a native context window of 1 million tokens, it understands text, images, audio, and video natively, capable of handling large document sets and diverse inputs.