What is Beam? The buzzy American AI model being called the 'DeepSeek of the West'
The company says it rivals some of China's leading open models while using far less computing power.
Reflection AI, a New York-based AI lab, has unveiled its first major model called Beam. The company is positioning Beam as a Western alternative to China's advanced open AI systems. Beam is an open-weight model designed specifically for coding, reasoning, and AI-agent tasks. This means developers can download the model's parameters, run them on their own hardware, and customize them for specific use cases, unlike models from companies like OpenAI and Anthropic which require their own APIs.
Beam boasts 501 billion total parameters but only activates 23 billion at a time. This design aims to make the model cheaper and faster to operate. Currently, Beam is undergoing final safety testing and will be available through a waitlist later this month, with the model's weights, a technical report, and developer tools to follow.
Misha Laskin and Ioannis Antonoglou, both former DeepMind researchers, founded Reflection in 2024. The startup has secured significant funding, including a $25 billion pre-money valuation in June, and is backed by investors such as Nvidia, Sequoia, and Citigroup.
Reflection claims Beam competes with models like Z.ai's GLM-5.2 and approaches Alibaba's Qwen 3.8-Max in performance, although it lags behind Moonshot AI's Kimi K3. The company emphasizes the efficiency of Beam, stating it can deliver results comparable to GLM-5.2 on advanced reasoning benchmarks while using three to four times less inference compute.
Beam's entrance into the market is seen by many as a significant response from the US to China's dominance in open AI models, especially given comparisons to China's DeepSeek.
Written by urgent.news from Business Insider's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
