Trends
Track how model card documentation and evaluation disclosure have evolved across AI labs.
51
Model Cards Tracked
475k
Longest Card (Anthropic)
6
Labs with Evals
1279
Total Evals Extracted
Card Length Over Time
How model card length has changed across successive releases for each lab. Cards are ordered chronologically within each lab — click a lab to isolate its trend line.
Anthropic
Google DeepMind
Meta AI
Mistral AI
OpenAI
xAI
Evaluation Topics per Model Card
What topics each model card reports evaluations on, and how that coverage evolves across releases. Stacked bars break down eval count by category — reasoning, coding, safety, math, and more.