Beam
Reflection's 501B-parameter open-weight MoE model for coding and agents
reflection.ai/blog/introducing-beam
Community-submitted · content unverified. This organization has not verified ownership or the information in this profile.
About
Beam is Reflection AI's first open-weight model, announced October 5, 2026: a sparse Mixture-of-Experts model with 501 billion total and 23 billion active parameters, built for coding, reasoning and agentic workloads. It was pretrained on 23.8 trillion tokens and then trained with large-scale reinforcement learning on about 10.5K NVIDIA GB300 GPUs. Beam is in early access, with weights, a technical report, model card and developer artifacts slated for release later in October 2026.
- Added via
- web
- Ownership
- unclaimed
Where to go
Product details
- Product kind
- Open Source
- Open source
- Yes
- Launched
- 2026