LLM & Text Generation Machine Learning11 min reading time

[AINews] Reflection Beam - 501B-A23B American Open Model

Latent.SpacePublished 20 days ago

Reflection has launched Beam, a US-trained, text-only 501B-parameter MoE model focused on coding, agentic, and scientific tasks. The model was trained from scratch on 23.8 trillion tokens and promises high efficiency and open weights under Apache 2.0 soon.

[AINews] Reflection Beam - 501B-A23B American Open Model

More on this story

More in LLM & Text Generation

PyTorch Conference 2026: Google and Meta Open-Source TorchTPU for Native PyTorch on TPUs

Nanbeige Releases Nanbeige4.3-33B, a 33B Open Agentic Coding Model

Researchers from Stanford and UC Berkeley Introduce TokenGate: Token-Gated Attention Makes Long-Context Prefill Up to 11.3x Faster