No offense to the team at Reflection AI, but Beam is an absolute flop
Literally, the only benefit you can say is we aren’t China
Outside of that, beam is mogged by like every lab
Introducing Beam: a highly efficient agentic open model with 501B total parameters and 23B active.
- Frontier reasoning efficiency
- Advances the Western open frontier on coding & agentic tasks
- Trained end-to-end from scratch
Full weights release this month.
Learn more about Beam: https://t.co/c3Qx2cpM8G