Defined term
Model card
Documentation describing a model's intended use, limitations, evaluation, and risks.
A model card publishes the intended use cases, training data summary, evaluation results, known failure modes, and ethical considerations of a model. For production deployments, we maintain internal model cards for every prompt + model combination shipped, so any reviewer can quickly understand what is deployed and why.
When it matters
When more than one prompt+model combination is deployed and you need a quick way for any reviewer to understand what is in production. Required artefact in regulated industries.
Real example
An internal model card for the 'lead-qualifier-v3.2' deployment: model (claude-sonnet-4-6), prompt version, intended use, last eval results, known failure modes, escalation contact. One-page format, queryable by dashboard, version-controlled in Git.
KPIs to watch
Model card coverage (100% of production prompts), refresh cycle (<30 days from prompt change), reviewer satisfaction score (>4/5 on usefulness).
Related terms
AI governance
Policies, processes, and controls that make an AI system auditable and accountable.
Audit log
Tamper-evident record of every model input, output, version, and reviewer action.
Grounding
Anchoring model output to verifiable source material to reduce hallucination.
Hallucination
Plausible but factually incorrect output generated by an LLM with no grounding.
See it in action
We use this every week
Book a 30-min call and we'll walk you through how Model card shows up in a real engagement we're running.
Book a 30-min call