Nvidia-backed Reflection AI launches Beam to challenge Chinese models

Nvidia-backed startup Reflection AI has launched its first open-weight AI model, named Beam, positioning itself to compete with lower-cost Chinese alternatives. The model is designed to challenge offerings like DeepSeek and Kimi in coding and agentic tasks. Reflection claims Beam is competitive with Z.ai’s GLM-5.2 and is approaching the performance of Qwen3.8-Max in these areas.

Beam boasts 501 billion total parameters but activates only 23 billion for each task, making it faster and more cost-effective by utilizing only part of its network. In comparison, Z.ai’s GLM-5.2 has 744 billion total parameters with 40 billion active parameters. This launch comes as U.S. tech firms aim to counter the rising threat of Chinese open-weight models, which are more affordable, customizable, and nearly as capable as leading models from OpenAI and Anthropic in generating code.

Reflection AI was founded in 2024 by former DeepMind researchers Misha Laskin and Ioannis Antonoglou. The company specializes in developing tools to automate software development, a rapidly growing application for AI. Earlier this year, Reflection secured a deal with SpaceX for additional computing capacity at the Elon Musk-led company’s Colossus 2 data center.