Reflection AI Unveils Beam: 501B-Parameter Open Agentic Model With 23B Active, Weights This Month
seanmcdonaldxyz · x · 2026-10-06
Reflection AI announced Beam, an efficient agentic open model with 501B total parameters and 23B active (MoE), claiming frontier reasoning efficiency and advancing the Western open frontier on coding and agentic tasks. Trained end-to-end from scratch, with full weights release planned this month. Observers are watching expert routing, VRAM/node requirements, quantization, sustained throughput, 1M-context degradation, tool-call reliability, licensing, and benchmark reproducibility.
More from Models
- 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
- Model Fight Card 16: Tenet models dominate jailbreak arena with 21/24 escapes — Dave_Maynor · 2026-11-20
- Coding agent index shows max effort costs 6x for fewer points; Gemini 4 Neon wins on value — Arindam_1729 · 2026-11-20
- Reddit: Claude Max for $29 Instead of $200 — How Do Resellers Profit? — mehdiweb · 2026-11-20