Nvidia-backed US startup Reflection AI unveiled Beam on October 5, its first open-weight artificial intelligence model. The company says Beam is a sparse mixture-of-experts model with 501 billion total parameters and about 23 billion active at a time.
According to Reflection's official announcement, the model is designed for coding, reasoning and agentic workloads. The company says it was pretrained on 23.8 trillion tokens and then subjected to large-scale reinforcement learning.
Reuters reported that Reflection is positioning Beam as a US alternative to lower-cost Chinese open models. Reflection says the model is competitive with Z.ai's GLM-5.2 and is approaching Alibaba's Qwen 3.8-Max on some benchmarks. Those performance comparisons are company-reported results.
Reflection says one large reinforcement-learning run generated more than 100 million rollouts using 10,500 Nvidia GB300 GPUs over four weeks. Beam is still undergoing final red-teaming and evaluations, with broader access to the model weights planned later in October.