Sovereign and Open-Weight AI Moves Toward Government Deployment

Aleph Alpha's Kolibri remains a reported bilingual German/English open-weight MoE aimed at sovereign and regulated deployments, while Reflection's newly announced Beam broadens open-weight competition with a 501B-total/23B-active MoE and unusually large RL/OPD rollout claims. Beam's weights, benchmarks, licensing, and training details are still needed; current capability claims are announcement-level.