MLCommons is a global, community-driven organization that brings together companies, universities, researchers, and nonprofits to accelerate AI innovation for the benefit of society. Guided by open collaboration and collective engineering, it develops tools and shared resources to help the AI community measure and improve the accuracy, safety, speed, and efficiency of AI systems.

Its offerings include industry-standard benchmark suites such as MLPerf, covering AI training, inference, storage, automotive, mobile, and other areas. These benchmarks provide consistent performance measurements that help engineers evaluate systems and guide responsible development. Through its AI Risk & Reliability work, MLCommons also supports a harmonized approach to assessing AI safety and reliability.

The organization builds and provides open, diverse datasets, data standards, research infrastructure, and tools for reproducible AI work. Its projects include Croissant, a metadata vocabulary designed to make machine-learning datasets easier to describe and use. Supported by more than 125 members and affiliates, MLCommons connects industry and academia through working groups, shared research, and community participation.