Reflection AI open-sources Beam, a 501B-A23B MoE trained from scratch under Apache 2.0
OfirPress · x · 2026-10-06
Reflection AI has released its first in-house model, Beam, a text-only 501B total / 23B active MoE, openly licensed under Apache 2.0, with full weights coming this month.
- The team says Beam advances the Western open frontier on coding and agentic tasks with frontier reasoning efficiency, trained end-to-end from scratch.
- They spent the past year scaling a production RL/OPD stack: a stable run on 10K GB300s produced over 100M rollouts across 1M tasks, significantly boosting final performance.
Related event: Reflection AI Unveils Beam, Its First Open-Weight 501B Model(18 posts)→
More from Infra
- Modal Runtime keynote ships Endpoint Candidates, Clusters and VM Sandboxes — graceisford · 2026-10-06
- Dev shelving Cloudflare K2 support: no event push to Workers, API not in spec — samgoodwin89 · 2026-10-06
- PyTorch Consolidates Media Stack: TorchCodec Becomes Single Home for Decode/Encode — PyTorch · 2026-10-06
- AI-generated key-value stores beat general-purpose DBs via specialized architecture — CShorten30 · 2026-10-06
- Only 30-40k robots installed in the US last year — the case for self-replicating factories — ihorbeaver · 2026-10-06
- Reflection says Beam hit ~19% MFU with 92% goodput during its 4-week RL run — alexpolozov · 2026-10-06