mot

name	mot
description	Evaluate and classify machine learning models based on the Model Openness Framework (MOF), and extract model metadata such as architecture, origin, producer, and components.
license	MIT

Skill: Model Openness Tool (MOT)

MOT provides a list of all the models currently registered and how they rank against the MOF classes.

Core Process

1. Locate Models: Identify models within the target repository (https://github.com/lfai/model_openness_tool/tree/main/models).
2. Extract Metadata: Parse model files to extract architecture, origin, producer, type, date of release, framework, and components.
3. Classify Openness: Evaluate the model against the Model Openness Framework (MOF) criteria based on its components.
4. Format Output: Present the classification and metadata in a structured, concise format.

Core Principles

• Direct Extraction: Fetch and parse raw metadata directly from the source JSON/YAML files.
• Strict MOF Adherence: Always map components to the official MOF evaluation criteria for accurate classification.
• Concise Reporting: Return only the requested fields (e.g., framework, architecture, producer) and the final MOF classification.

When to Use

• When finding a list of open models or specific model metadata from the Model Openness Tool repository.
• When classifying a machine learning model's degree of openness using MOF.
• When extracting specific model properties like architecture, release date, or components.

When Not to Use

• When searching for general-purpose software libraries or non-AI code repositories (use github-topics instead).
• For evaluating the runtime performance, latency, or accuracy benchmarks of an LLM.
• If you need to actually download and run the weights of a machine learning model.

Common Pitfalls

• Ignoring the Framework: Classifying a model simply as "open source" without actually evaluating it against the specific tiers and definitions of the Model Openness Framework (MOF).
• Outdated Metadata: Relying on external web searches for model licensing instead of parsing the official YAML files maintained within the MOT repository.
• Hallucinating Components: Assuming a model includes training data or specific architecture details without explicitly extracting that information from its metadata file.

Commands / Usage Patterns

Fetch and parse a model definition directly from the source repository using curl and yq:

curl -s https://raw.githubusercontent.com/lfai/model_openness_tool/main/models/<model-file>.yaml | yq '.'

References

• List of Models (Name, Organization, Classification)
• Licenses
• Model Openness Tool Repository
• Models Directory