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.
More from Infra
- Nvidia down ~40% from October peak as Anthropic IPO pops 34% above expectations — The Verge AI · 2026-11-20
- Ion Stoica explains why complex AI workloads increasingly fit Ray's design — PyTorch · 2026-11-20
- GPU rents up, token prices down: inference providers' break-even load jumps 22%–75% in 5 months — AccBalanced · 2026-11-20
- Missing config file was throttling GPUDirect Storage — fix yields 3x reads, 9x writes — AccBalanced · 2026-11-20
- NVIDIA Dynamo cuts RL rollout bottlenecks with KV-aware routing and delta weight updates — NVIDIAAI · 2026-11-20
- Arm reaches the AI laptop: Surface Laptop Ultra runs 120B models locally — kimmonismus · 2026-11-20