Sources: project SGLang spins out as RadixArk with $400M valuation as inference market explodes

RadixArk: Transition from Open-Source Tool to Commercial Startup

A portion of the team behind SGLang, a widely-used open-source tool in the AI model training space, has shifted focus to their newly established commercial venture, RadixArk. This startup, which was officially unveiled last August, is making waves in the industry.

Originally conceived as SGLang within the confines of UC Berkeley’s Databricks lab under the guidance of Ion Stoica, RadixArk has recently achieved a valuation of approximately $400 million following a funding round led by Accel, as per sources familiar with the matter.

Prior to this, RadixArk had garnered angel investment from notable figures like Intel CEO Lip-Bu Tan, signaling early confidence in its potential.

Ying Sheng, formerly associated with xAI and a key contributor to SGLang, has assumed the role of co-founder and CEO at RadixArk. Her decision to depart from Elon Musk’s AI venture to spearhead this new initiative underscores the promise RadixArk holds in the AI ecosystem.

RadixArk, alongside SGLang, is dedicated to refining inference processing efficiency, a critical aspect of AI operations. By optimizing this process, significant cost savings can be realized, making it a compelling proposition for businesses.

In a parallel development, vLLM, another project focusing on inference optimization, has also transitioned into a startup, with discussions underway for substantial funding that could propel its valuation to $1 billion, according to reports.

The investment landscape in this domain is vibrant, with Andreessen Horowitz reportedly leading the funding charge for vLLM. However, specifics of this investment are yet to be confirmed, underscoring the dynamic nature of the AI sector.

Both SGLang and vLLM have roots in Ion Stoica’s lab at UC Berkeley, a hotbed of innovation in the AI realm. Stoica’s pivotal role in nurturing these projects further solidifies their credibility in the market.

RadixArk’s commitment to advancing AI capabilities is evident through initiatives like the open-source AI model engine SGLang and the development of Miles, a specialized framework for reinforcement learning. These endeavors aim to empower businesses in their AI endeavors.

While RadixArk continues to offer many tools for free, the introduction of hosting service fees reflects the maturation of its business model, aligning with industry trends.

The surge in funding for startups specializing in inference infrastructure highlights the growing importance of this segment within the AI landscape. Notable players like Baseten and Fireworks AI have made significant strides in this space, signaling a competitive and dynamic market environment.