The Arena Models endpoint returns all AI models discovered and tested by Cranium across your tenant. Use this endpoint as a model catalog when correlating model identifiers across endpoints, and as a reference for model metadata when integrating Arena data with downstream systems.

List Arena Models

Returns a paginated list of models cataloged in the Arena.

Request:

GET /api/public/arena/models

Authorization:

Bearer token. See Authentication & Generating Credentials.

Query parameters:

Response record:

json

{
  "modelId": "guid",
  "modelName": "llama-3-8b-instruct",
  "modelFamilyId": "guid",
  "modelFamilyName": "Llama 3",
  "modelSource": "HuggingFace | Internal",
  "modelType": "LLM | DL | TradML",
  "referenceUrl": "https://huggingface.co/...",
  "modelStatus": "string",
  "severity": "Critical | High | Medium | Low | None",
  "averageAttackSuccessRatio": 0.42,
  "attackCategoriesVulnerableCount": 3,
  "weaknessCount": 5,
  "knownRisksCount": 2,
  "automatedTestCount": 10,
  "manualTestCount": 1,
  "aiSystemCount": 2,
  "penTestCompletedAt": "2026-03-15T00:00:00Z | null",
  "discoveredAt": "2026-01-05T00:00:00Z",
  "updatedAt": "2026-04-10T00:00:00Z"
}

Field Definitions

• modelId: Unique identifier for the model. Stable across updates. Use this value to correlate with modelId on the Arena Model Vulnerabilities endpoint.
• modelName: Human-readable name of the model as displayed in the Arena.
• modelFamilyId: Identifier of the model family. Models that share a base architecture (for example, multiple Llama 3 variants) share a modelFamilyId.
• modelFamilyName: Display name of the model family.
• modelSource: Origin of the model. HuggingFace for models cataloged in HuggingFace Hub, Internal for organization-specific models.
• modelType: AI type classification. LLM for large language models, DL for deep learning models, TradML for traditional machine learning models.
• referenceUrl: URL to the model's reference page (for example, the HuggingFace model card).
• modelStatus: Current status of the model. One of Untested, Queued, InTesting, Completed, Failed, NotTestable.
• severity: Highest severity classification across the model's tested attack categories. One of Critical, High, Medium, Low, None.
• averageAttackSuccessRatio: Average ratio of successful attacks across all tested attack categories. Expressed as a decimal between 0 and 1. A value of 0.42 indicates that 42 percent of attack attempts succeeded on average.
• attackCategoriesVulnerableCount: Number of attack categories in which the model was found vulnerable.
• weaknessCount: Number of identified weaknesses associated with the model.
• knownRisksCount: Number of known risks documented for the model.
• automatedTestCount: Number of automated penetration tests run against the model.
• manualTestCount: Number of manual penetration tests run against the model.
• aiSystemCount: Number of AI Systems in your tenant that include this model.
• penTestCompletedAt: Timestamp at which the most recent penetration test completed. Null when no test has been run. ISO 8601 format.
• discoveredAt: Timestamp at which the model was first cataloged in the Arena. ISO 8601 format.
• updatedAt: Timestamp at which the record was last updated. ISO 8601 format. This is the field the sync cursor tracks.