Building the Foundations
Accelerating AI and Machine Learning with Open Standards
The speed of innovation within the AI and machine learning ecosystem is faster than ever before, and shows no signs of slowing with new architectures regularly hitting the headlines. While such innovations are fantastic in driving performance and functionality, the sheer diversity of hardware and software solutions poses a challenge for developers hoping to support the latest advancements, while also reaching the widest audience. Systems built for AI and machine learning typically base their performance on custom processors, CPUs, GPUs or FPGAs. The Khronos Group is a recognized expert in creating open standards that connect software to silicon, and to help coordinate our AI and ML activities the Khronos Group has formed the ML Council.
Driving ML Open Standards
About the Khronos Machine Learning Council
The Khronos ML Council is a collaborative forum composed of Khronos member companies from across the technology industry whose ambitious goal is to ensure that open, interoperable standards evolve in line with industry needs, enabling broader adoption and more efficient innovation in the AI and ML ecosystem. The forums activities include:
- Discussing the challenges in cross-platform AI/ML development
- Promoting the exchange of knowledge in the AI/ML domain
- Building industry consensus on effective AI/ML acceleration directions
- Providing input to the evolution of existing open API standards
- Providing feedback to Khronos members and API implementers
- Conduction market research to better inform decision making
At SIGGRAPH 2025 we explored the problem of why the AI acceleration ecosystem is so complex, why AI performance portability is so hard, as well as discussing where open standards offer a solution.
Ecosystem Research
The Khronos AI Ecosystem Research Project
Khronos APIs are already used by many ML runtimes and frameworks, but there’s always more we can do to streamline the ecosystem. To guide our next steps, we are undertaking the AI Ecosystem Research Project to obtain real-world feedback to better understand the hardware and software ecosystem for machine learning and the role and opportunity for standards bodies such as the Khronos Group to advance pervasive acceleration of machine learning.
Preliminary results from this project are already available and we would like to thank to Futurewei for sponsoring this project.
View Research Project Update - Page 9 Subscribe for Final ReportOpen Standards
Machine Learning Standards from Khronos
Khronos is home to some of the world’s leading standards for heterogeneous compute acceleration for embedded devices, desktops, and large HPC servers, and is driving the evolution of these open standards to help developers deploy accelerated AI and ML algorithms and applications efficiently across multiple platforms. |
Exchange Format for connecting trained networks with inference engines. | Learn More |
 |
Low-level parallel programming of heterogeneous processors. | Learn More |
 |
Computer vision acceleration API with ML extension and tensor objects. | Learn More |
 |
Shading language and compiler with ML capabilities. | Learn More |
 |
The standard IR for parallel compute and graphics. | Learn More |
 |
Single source C++ abstraction layer for heterogeneous processors. | Learn More |
 |
Low-level, low-overhead cross-platform GPU API. | Learn More |
Help Drive Open Standards for AI and ML
Join the Khronos Group
If you are interested in the future of ML acceleration, consider becoming a Khronos Member and joining the Machine Learning Council. Khronos membership gives you the opportunity to collaborate with leading companies, contribute to the development of open standards, and help shape the direction of AI and ML technologies. Khronos offers multiple levels of membership designed to suit organizations of all types and sizes, making it easy for start-ups, research groups, and global enterprises alike to get involved and have a voice in driving the evolution of the ML ecosystem. Khronos membership entitles you to join any of the standards Working Groups - from OpenCL and SYCL to Slang and Vulkan.