Beam 1.0 501B A23B is a sparse Mixture-of-Experts text model designed for token-efficient reasoning, software engineering, agentic tool use, and long-horizon task completion. It carries 501B total parameters with 23B active, built on a 52-layer architecture with 5,120 hidden size, 48 attention heads, and 8 KV heads. Gated grouped-query attention pairs 43 sliding-window layers with 9 full attention layers, and the 49 MoE layers route each token to 8 of 256 experts. Supporting a 1M-token context, it is refined through reinforcement learning across coding, agentic, and STEM environments.
Model specifications and capabilities are published by the model author and reproduced here from HF Model Card (reflection/Beam-1.0-501B-A23B). Released under Apache 2.0. Further reading: Blog
Choose your hardware and inference engine to get deployment commands and performance benchmarks tailored to your infrastructure
1,440 GB
Enter your email to get access to this content
Benchmarks measured by Vultr on B200 · vLLM. Throughput and latency vary with concurrency, input length and engine configuration, so treat these as a comparison baseline, not a service guarantee.