Intel Corporation (NASDAQ: INTC) has partnered with RadixArk to optimize SGLang, an open-source AI inference framework, for Intel platforms. The collaboration aims to help developers and enterprises deploy large AI models and build applications with improved performance, lower latency and greater cost efficiency.
The companies plan to integrate optimizations directly into SGLang, enabling developers to run AI inference workloads across Intel central processing units (CPUs), graphics processing units (GPUs) and accelerators. The effort will focus on improving software compatibility and performance while preserving the flexibility of open-source development.
Intel said the partnership addresses the growing need for production-ready AI infrastructure as enterprises move from evaluating individual model benchmarks to deploying complete AI systems.
“Enterprise AI is moving beyond isolated performance benchmarks to solving full-system production deployments,” said Anil Nanduri, vice president of AI Products and Go-to-Market at Intel’s Data Center Group.
He said the collaboration would provide developers with Day-0 readiness and optimized inference software across Intel hardware, while allowing them to choose the frameworks and technologies that suit their requirements.
Optimizing AI Inference for Enterprise Workloads
As AI models become larger and more computationally demanding, enterprises increasingly need infrastructure that can distribute workloads across different types of computing resources. Inference systems must match incoming requests with appropriate models and hardware while balancing performance and operating costs.
Open inference frameworks play a role in coordinating these workloads across heterogeneous computing environments. The Intel-RadixArk collaboration seeks to strengthen this software layer by incorporating platform-specific improvements into SGLang.
For developers, the integration is expected to support deployment of leading AI models on Intel hardware using established open frameworks and interfaces. It will also make use of Intel hardware capabilities, including Intel Advanced Matrix Extensions (AMX), which are designed to accelerate matrix operations used in AI and other compute-intensive workloads.
The companies said these optimizations could help reduce inference latency, improve time to production and give organizations additional options for using existing data center and cloud infrastructure.
Ying Sheng, CEO of RadixArk, described the partnership as a step toward making high-performance AI inference more accessible through open-source software.
“By upstreaming deep optimizations into the core framework, we are eliminating constraints and ensuring developers never have to compromise between open standards and high performance,” Sheng said.
Open-Source Development and Production AI
The collaboration reflects Intel’s strategy of contributing hardware optimizations directly to widely used open-source frameworks rather than limiting improvements to proprietary software environments.
Upstreaming these changes into SGLang’s core framework could make the optimizations available to a broader developer community and simplify deployment across supported Intel platforms.
The companies did not disclose a timeline for the optimizations or provide specific performance benchmarks. Their stated objective is to make open-source inference software more capable of supporting scalable enterprise AI applications while retaining hardware and framework choice.
