modelami

Modelami is a collaborative project and catalog designed to document and compare machine learning models across domains. The project provides a centralized registry where researchers and developers can submit model cards, evaluation results, licenses, and intended uses. Modelami emphasizes reproducibility, transparency, and ethical considerations by encouraging standardized documentation and open benchmarks.

The project was launched in 2022 by a coalition of universities, research institutes, and open-source communities. It grew through community contributions and partnerships with AI labs and industry groups, and it aims to become a common reference point for model documentation.

Key features include a model registry with versioning, metadata fields such as task, architecture, training data, license, biases, and deployment constraints, as well as an evaluation harness with standardized test suites. The platform provides APIs for programmatic access and supports interoperability with popular ML frameworks and packaging formats. Governance policies and a discussion forum facilitate community oversight and knowledge sharing.

Modelami is used by researchers to compare models for literature reviews, by practitioners to inform tool selection for production, and by educators to teach model documentation. While it can reduce duplication and encourage accountability, critics warn that benchmarks may not capture real-world performance and that sensitive data considerations must be managed carefully.

See also: Model card, ML model registry, AI transparency initiatives.