英特尔联手 RadixArk 优化开源 SGLang 推理栈：让领先模型在英特尔平台首日可用
总部位于加州 Santa Clara 的英特尔公司11月12日宣布，与 RadixArk 合作优化开源推理栈 SGLang，使其适配英特尔平台，帮助开发者与企业更高效、更低成本地部署领先的超级智能（SI）模型、构建相关应用。
英特尔数据中心集团负责超级智能产品与市场推广的副总裁 Anil Nanduri 说，企业级超级智能正从孤立的性能跑分转向解决全系统的生产部署问题；这次合作让开发者在第一天就获得直接为英特尔中央处理器、图形处理器与加速器优化过的高性能推理软件，并能以开放方式自由选择最适合自己的工具、框架与技术。
合作将扩大开发者在经企业验证的英特尔硬件上部署领先模型的选择。模型开发者与软件工程师可以在标准、开放的框架与接口上构建，同时用足英特尔的计算密度与内置硬件加速功能，包括 Intel AMX（高级矩阵扩展）；开发者由此获得更快的投产速度、更低的延迟，以及跨现有数据中心与云容量的灵活部署选项。
RadixArk 首席执行官 Ying Sheng 说，与英特尔的合作是开源推理的一个重要里程碑，让 SGLang 的优化直接进入英特尔平台；把深度优化回馈到核心框架，可以消除约束，让开发者不必在开放标准与高性能之间取舍。
英特尔说，把硬件与软件优化结合起来，体现了该公司把改进直接贡献给社区软件的做法，目标是为把突破性模型变成可扩展的生产级应用提供高性能路径，同时保留灵活性、选择与架构自由。
（来源：英特尔）
Intel Teams With RadixArk to Optimize Open-Source SGLang Inference Stack for Day-0 Readiness on Intel Platforms
Intel Corporation, headquartered in Santa Clara, California, announced on November 12 a collaboration with RadixArk to optimize SGLang, an open-source inference stack, for Intel platforms, aiming to help developers and enterprises deploy leading super intelligence (SI) models and build applications more efficiently and cost-effectively.
Anil Nanduri, vice president for SI products and go-to-market at Intel's Data Center Group, said enterprise SI is moving beyond isolated performance benchmarks to solving full-system production deployments, and that the collaboration gives developers Day-0 readiness with high-performance inference software directly optimized for Intel CPUs, GPUs and accelerators in an open approach, with freedom to choose the tools, frameworks and technologies that best fit their needs.
The collaboration expands developer choice to deploy leading models on enterprise-proven Intel hardware. Model builders and software engineers can build on standard, open frameworks and interfaces while taking full advantage of Intel's compute density and built-in hardware acceleration, including Intel Advanced Matrix Extensions (Intel AMX), gaining faster time-to-production, reduced latency and flexible deployment across existing data center and cloud capacity.
Ying Sheng, chief executive of RadixArk, said the collaboration with Intel is a major milestone for open-source inference, enabling SGLang optimization directly on Intel's platforms, and that upstreaming deep optimizations into the core framework eliminates constraints so developers never have to compromise between open standards and high performance.
Intel said bringing hardware and software optimization together reflects its approach of contributing improvements directly to community software, aiming to provide high-performance paths for turning breakthrough models into scalable, production-grade applications while preserving flexibility, choice and architectural freedom.
(Source: Intel Corporation)
来源：英特尔
发布：GSM·美国（发布链接）
图片来源：图：英特尔位于加州 Santa Clara 的总部大楼。Sixflashphoto 摄，CC BY-SA 4.0，经 Wikimedia Commons。 / Photo: Intel's headquarters building in Santa Clara, California. Photo by Sixflashphoto, CC BY-SA 4.0, via Wikimedia Commons.
GSM Global · 本篇英文版 English edition of this story
