The open-source reasoning engine SGLang team has completed a $100 million seed round financing, with Accel leading the round.

According to Dongcha Beating monitoring, the team behind the open-source inference engine SGLang has officially established the company RadixArk, completed a $100 million seed round financing, and achieved a post-money valuation of $4 billion. Accel led the round, with participation from Spark Capital, NVIDIA (through NVentures), and AMD, two chip rivals, as well as from MediaTek.

SGLang serves as an intermediate layer between models and hardware, reducing AI inference costs by more efficiently scheduling GPU memory. The project, launched in 2023, currently processes trillions of tokens daily for Google, Microsoft, NVIDIA, Oracle, xAI, and others. RadixArk also operates another open-source project, Miles, focusing on large-scale reinforcement learning training, which has been used for training Mixture of Experts (MoE), a model architecture that activates only a subset of parameters to reduce computational overhead.

The company was founded by Banghua Zhu and Ying Sheng. The financing will be used to expand the two open-source projects, support more model architectures and hardware platforms, and launch enterprise-grade managed infrastructure services.