Reflection AI unveils Beam: 501B-param MoE open model with 23B active
alexpolozov · x · 2026-10-06
Reflection AI introduced Beam, its first model: a highly efficient agentic open MoE with 501B total / 23B active parameters, trained end-to-end from scratch, claiming frontier reasoning efficiency and advances on Western open-weight coding & agentic tasks, with full weights releasing this month. A pretraining team member describes building a frontier MoE from scratch in a small team — fitting scaling laws, taming residual norms, debugging loss spikes.
More from Models
- Jeremy Howard: Google shipped the best model ever made; OpenAI its worst, same day — hugobowne · 2026-11-20
- Polymarket puts 83% odds on Google having the best AI model by year-end — Polymarket · 2026-11-20
- One year later, model naming is still a mess: the gpt-5.1-codex-max anniversary — heyneighbor · 2026-11-20
- Developer slams Grok models as 'garbage', shares coding agent's repetitive trial-and-error log — feregri_no · 2026-11-20
- Drex Router cuts small-game build cost to $1.26 vs $13.44 on Claude Opus alone — testingcatalog · 2026-11-20
- OpenAI slammed for AI slop: 'Just hire photographers to RLHF imagegen for a day' — flowersslop · 2026-11-20