Frontier Model Reporting Tracker

Tracking — safety & transparency documents (system cards, model cards, frontier safety reports, transparency reports) across — models from — leading AI developers.

Last updated — · by Leon Staufer

Loading data…

Documents are published by the respective companies. Dates and links are gathered from official sources and may contain errors. Not all dates have been independently verified. Where a gap is marked ~ it is derived from published dates rather than independently verified.

This list may over-count model reports, as some documents lack specific safety-evaluation results. Model, product and inference-time scaffolding variants (e.g. a reasoning mode like Gemini Deep Think, or an agent built on a base model) get their own row even when not branded as a new model. They are credited with a model report only when a document reports evaluation results for that variant.