Reflection releases open MoE Beam-1.0: 501B total/23B active, 1M context, 3x fewer reasoning tokens
jacek2023 · reddit · 2026-10-16
Reflection has released Beam-1.0-501B-A23B as an open model on Hugging Face, with BF16, FP8 and NVFP4 checkpoints.
Architecture: sparse MoE with 501.3B total / 23.4B active parameters, 52 layers (3 dense + 49 MoE), 256 experts, Gated GQA + sliding window attention, 1M-token context, 152k vocab, June 30 2026 knowledge cutoff.
Claimed capabilities:
- Token-efficient reasoning: comparable quality to leading open models with 3x fewer inference-time reasoning tokens.
- Software engineering: resolves real GitHub issues, works in CLI/DevOps and agentic coding harnesses.
- Agentic tool use: long-horizon tasks, MCP integration, browsing.
- Math/science reasoning: graduate-level problems and theorem proving, transferring to debugging and self-correction.
More from Models
- Golden Gate Claude explainer resurfaces: interpretability premise still holds — HarperSCarroll · 2026-10-16
- Hugging Face livestreams release of a new open-weight model — huggingface · 2026-10-16
- Step 5 Preview hits #2 on OpenRouter weekly leaderboard with 19.5T tokens in one week — StepFun_ai · 2026-10-16
- Meta's open Muse Spark needs 493GB of memory — far beyond any consumer laptop — TejasKumar_ · 2026-10-16
- Apache 2.0 is the underrated part: Beam beats rival open drop on license — bigblueboo · 2026-10-16
- Apollo Research: GLM-5.3 spontaneously reasons about how its actions shape its future training data — Confident_Salt_8108 · 2026-10-16