Author: IBL News

IBL News | New York

Reflection AI, a New York AI startup that presents itself as the U.S.’s answer to open-source Chinese AI, released its first model, Beam, claiming it outperforms Western open models.

This company targets a large market of businesses, governments, and other entities building sovereign AI systems that don’t want to use Chinese models.

Founded in 2024 by former DeepMind researchers, Reflection AI said that the model is particularly strong at coding and agentic tasks. It also needs three to four times less computing power to reason, making it faster and cheaper to run.

The cofounder and CEO, Misha Laskin, said the model’s capabilities are comparable to China’s Z.ai’s recent model, GLM-5.2, and Alibaba’s most advanced model, Qwen 3.8-Max — which trails the top closed American models from OpenAI and Anthropic.

Several large Western companies have recently adopted Chinese models to reduce costs for certain repetitive tasks, such as customer service and writing routine code. Beijing is also promoting open source as a diplomatic tool to help other countries in their tech development.

However, because open models can be downloaded and run with fewer safeguards, their rise has also raised security concerns: US and UK government evaluators found that recent Chinese models could help hackers exploit code vulnerabilities.