Term
Model Card
A model card is a standardised datasheet for an AI model — documenting purpose, training data, performance, limitations and ethical considerations in one place.
Model card — explained in more detail
The concept goes back to the paper “Model Cards for Model Reporting” (Mitchell et al., Google, 2019) and has since become the de-facto standard. A typical model card describes: what the model does and what it is intended for, what data and methods were used to train it, what benchmark scores it achieves, what bias and safety analyses exist, what known weaknesses or out-of-scope applications apply, and which licence and responsible party are attached. On Hugging Face, the model card is simply the README.md of a model repository — many tools (inference servers, evaluation pipelines) read individual fields from it automatically.
Example / Practical context
A common workflow when picking a local LLM: open the candidate model on Hugging Face, read the model card — check the licence (commercial use allowed?), check the training cutoff (how current is the knowledge?), compare benchmark scores against your task, scan the limitations section for known weaknesses. Closed-source providers such as Anthropic and OpenAI publish similar documents for their models — usually called “model card” or “system card”.
Distinction from related concepts
A datasheet for datasets analogously documents a training dataset rather than a model. A system card (OpenAI/Anthropic) is an extended model card that also covers safety evaluations and red-team results. Datasheets, benchmark reports and licence files are separate documents — the model card bundles the most relevant information for users in one place.
Entdecke mehr
AI Workflows by Keyword: How We Make Recurring Routines Enforceable
A typed keyword triggers a fixed AI routine — and every single step must be committed before the next one appears. Why that's the actual trick.
GlossarReasoning Effort
Reasoning Effort is a control parameter on modern reasoning models that sets how much internal step-by-step thinking (thinking tokens) a model spends before answering — higher levels raise quality but also latency and cost.
LexikonLLM Hallucinations — Causes and Remedies
Why LLMs confidently invent falsehoods, what types of hallucinations exist, and which remedies actually help — RAG, source enforcement, verification.