RadixArk, the commercial company behind the open source inference engine SGLang, has been valued at roughly $400 million in a recent funding round led by Accel, according to people familiar with the deal. The startup, launched last August, grew out of research originally developed in 2023 at the UC Berkeley lab of Ion Stoica, and has since attracted adoption from companies including xAI and Cursor.
RadixArk is led by co-founder and chief executive Ying Sheng, a former xAI engineer and Databricks research scientist. The company focuses on optimizing AI inference workloads, enabling models to run faster and more efficiently on existing hardware. While continuing to maintain SGLang as an open source project, RadixArk is also developing Miles, a reinforcement learning framework, and has begun charging for hosting services. The rise of RadixArk reflects a broader shift as open source AI infrastructure projects evolve into venture-backed companies amid surging demand for inference optimization.
