Reflection AI unveils Beam, a 501B-parameter open-weight model

Reflection AI released Beam, its first open-weight sparse MoE model with 501B total and 23B active parameters, targeting coding, reasoning and agentic workloads. Pretraining on 23.8T tokens took under four weeks on 6,144 Nvidia GB300 NVL72 GPUs; early access is via waitlist.

• 501B total parameters, 23B active in a sparse MoE design
• Trained on 23.8T tokens in under 4 weeks on 6,144 GB300 NVL72 GPUs
• Scores: 80.9 SWEBench Verified, 80.1 Terminal Bench v2.1, 97.8 AIME 2026
• Uses 3–4x less inference compute than GLM-5.2 at comparable scores