ACFIJKLMNOPQRSTUVWXYZAAABAC
1
Model Card Metadata
2
CategoryData PointDescriptionFeasibilityComments/Questions
3
Model OverviewModel NameLists the internal technical name of the AI model.
4
Model Brand NameExternal or customer facing name for the AI model or system.
5
Model Release Date Lists the release date of the model.
6
Model DeveloperLists the individual or organization which developed or created, designed, and trained the AI model. This can also be known as the model author.
7
Model VersionLists the version of the model.
8
Model DescriptionProvides key information summarizing the model's architecture, functionality and capabilities, intended use, and limitations, enabling the user to determine whether the model is relevant for use.
9
SourceA URL or other path to where this model lives, such as a public repository.
10
Publisher Captures the entity that makes the model publicly available for use, download, or access. The publisher is often responsible for maintaining the model, ensuring documentation, and in certain cases, managing the model's licensing.
11
Provider DetailsCaptures the contact details of the provider who makes the AI model available for use as well as hosting and maintaining the model.
12
Model ProfileModel Format File format/type or standardized structure used to store and represent an AI model's architecture, parameters, weights, and metadata.
13
Framework or LibraryFramework and version used to develop, train, and deploy the machine learning model including libraries the model is dependent on.

Any libraries that the model is dependent on. For example, in HuggingFace, models may list dependencies on libraries like PyTorch, Transformers, ONNX, Diffusers, etc.
14
IndustrySpecifies the primary industry where the AI model is intended to be applied.
15
Language SupportLists the language(s) the model is trained on or supports.Format for language specififaction: ISO_639-1? (e.g. en, fr, de, ...)
16
Model Application/TaskDescribes the specific application or task the model is designed for.
17
Model TypeIdentifies the type of machine or deep learning model.
18
Model ParametersVariables or coefficients within a machine learning model that are learned or adjusted during training to optimize its performance.
19
Input(s)Specifies the type of data that the model takes as input to generate predictions or outputs. It describes the format and nature of the data that the model processes.
20
Output(s) Describes the type of data that the model produces in response to the inputs.
21
Intended UseDescription of the intended use case(s) of the model including the targeted audience or users, context, and outcomes. This includes describing usage of the model which constitutes misuse.
22
Intended DevelopersIdentifies who the model is designed to serve in terms of development expertise, including data scientists, ML engineers, or additional technical roles, providing context for model usage, maintenance requirements, and areas of responsibility.
23
Unintended UseA description of what use cases or usage of the model is not in scope for the model from a technical, ethical, and/or legal perspective.
24
Technical LimitationsDescription of any known technical limitations of the model including but not limited to: the factors that might degrade model performance and types of data model should be expected not to perform well on.
25
Citation DetailsProvides the preferred citation of the model typically provded by the developer.
26
Model ProvenanceBase ModelIf this model is built on a parent or pre-trained model, provide the link to the base model.
27
OriginLists the model's originating country.
28
Model Source TypeLists whether the mode was internally developed, open source, vendor-provided, pre-trained, third-party licensed model, pre-built platform model, hybrid (multiple sources) etc.
29
Developing OrganizationLists the organization which developed or created, designed, and trained the AI model.
30
Change LogLog of all changes and updates made in each version of the model.
31
Model ArchitectureArchitecture Family
Broad category of machine learning models to which the model belongs based on high-level model architecture such as design patterns or techniques.
32
Network Architecture Specific classification within the architecture family based on the overall structure and organization of a model's components.
33
Model Architecture TypeDescription of the structure and design of the model including layers, connections, and components.
34
Model License(s)The list of software licenses and associated links for the model.
35
Model Training Data and EvaluationTraining Data License(s)The list of licenses governing the use of teaining data for the model.
36
Training DataDescription of the training data sets utilized to train and validate the model including the name, source (URL), and version of each datatest, type of data (such as first party, public, or third party), and license information.
37
Training Data ParametersTraining datasets used to develop and validate the model include details on data sources, types of data collected, collection methods, dataset size, quality, provenance, class or label distributions, and preprocessing techniques applied.
38
Training Procedures Lists the training procedures including preprocessing steps, training hyperparameters, as well as speeds, sizes, and times.
39
Test DataDesription of what is being measured as part of model evaluation and the data sets used to evaluate model performance, data sources, data size, data distribution, and preprocessing steps applied to the test data as well as reasoning for selection of the data sets.
40
Testing MethodologyDescribes the testing methodology employed.
41
Test Data and Performance Metrics Lists the quantitative metrics used for evaluation of model performance based on test data.
42
Test Data Results SummaryOverview of the model's performance on test data including results, limitations, and improvements.
43
Model Performance Model StatusCurrent status of the AI/ML model within the AI system, reflecting model maintenance and performance levels.
44
End of Life Date (EOL Date)Captures date when AI model will no longer be supported.
45
End of Life Date ReasoningProvides the reason for the model's end of life.
46
Licensing RestrictionsExplain where or how the model can and cannot be used including examples of potential risk scenarios.
47
Model TransparencyEthical LicenseLists the applicable ethical licenses for the AI model that imposes restrictions on its use, grounded in ethical considerations.
48
Encryption TechniquesTechnique used to protect the model's intellectual property and training data from unauthorized access and tampering.
49
Certificate Data MetadataThe type, number and expiry date of the certificate issued by the notified body and the name or identification number of that notified body.
50
Certificate Data Summary of key data elements of the AI system and the model's compliance with specific regulations and standards.
51
Potential Known Risks and HarmsList of potential risks and harms associated with the model, including but not limited to: data privacy, security risk, environmental impact, bias and discrimination, as well as unintended consequences.
52
Model Signature(s)Attestation details of model signing including model signing provider, date, and time.
53
Sensitive Data Handling MeasuresStates whether Personally Identifiable Information (PII) or sensitive data were used in model training, data handling practices, and security measures employed in the use of the data.
54
Security ConsiderationsWarnings regarding potential security concerns associated with model usage, along with mitigation strategies implemented to address these risks.
55
Bias and Fairness EvaluationExplanation of known biases or limitations in fairness in model based on the identification of protected attributes such as race, gender, age, disability status, and note if they were included/excluded in the model and/or training data as well as mitigation strategies implemented.
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100