What happened
On October 16, 2026, Hugging Face said Beam, the first model from US lab Reflection, is now on its platform. It ships as open weights under the Apache 2.0 license. Users can reach it through Hugging Face Inference Providers, with Baseten, Fireworks, and Together supplying the backend, and more providers are expected shortly.
Why it matters
Developers can tap Beam, a mixture-of-experts system with 501 billion total parameters, 23 billion engaged per token, and a 1-million-token context window, via Hugging Face. Reflection also intends to award $10,000 in Hub credits to top projects built on Beam.
Read the original at Hugging Face