Technology ❯ Artificial Intelligence ❯ Machine Learning Architectures ❯ Mixture-of-Experts Models

Beam Sparse MoE with 501 Billion Parameters

Reflection Unveils Beam, a 501B-Parameter Open-Weight Model

Reflection says Beam delivers coding and reasoning performance on par with top Chinese open models while using far less inference compute, with full weights due for public release later in October.