Reflection says Beam hit ~19% MFU with 92% goodput during its 4-week RL run
alexpolozov · x · 2026-10-06
Responding to outside estimates of Beam's training efficiency, Reflection's alexpolozov said actual MFU was slightly above 19% — outside assumptions were off — and per the official post they hit 92% goodput toward the end of the sub-4-week run, with rewinds included.
He acknowledged room to reach industry SOTA, noting their next in-progress model already achieves higher MFU, and stressed starting pretraining early was crucial so the RL team had time to iterate. A tech report is coming later this month.
More from Infra
- SpaceX files for 32.4-mile Florida gas pipeline to make Starship methane on-site — DimaZeniuk · 2026-10-06
- Inference Companies Ship Gateways, Gateway Companies Ship Inference — the Lines Are Blurring — michellechen · 2026-10-06
- Can an 8GB Radeon 5700 Run Image-to-Video? ComfyUI Low-VRAM Question — Fluffy-Composer9675 · 2026-10-06
- Pairing a Blackwell RTX 4000 24GB with an Old RTX 3060 12GB for Local LLMs — Otherwise-Tangelo-52 · 2026-10-06
- Local LLM for VFX: 2x DGX Spark vs M5 Ultra 256GB for Blender/Houdini via MCP — abrasmel · 2026-10-06
- Modal Runtime keynote ships Endpoint Candidates, Clusters and VM Sandboxes — graceisford · 2026-10-06