Reflection AI

Open-weight large language models for enterprise and government customers, available through an API or deployable in private cloud, air-gapped and edge environments. Models include Beam, a sparse mixture-of-experts model with about 501 billion total parameters.