Guide: Model Cards

Document your models: versions, hashes, metrics, schemas, intended use, limitations.

Register a Model

model_card.pypython

model = client.models.create(
  model_id="credit-model-v2",
  version="2.1.0",
  model_hash="sha256:9f86d081884c7d659a2feaa0c55ad015a3bf4f1b2b0b822cd15d6c15b0f00a08",
  training_date="2025-01-01",
  framework="scikit-learn",
  performance={"accuracy": 0.94, "precision": 0.91, "recall": 0.89, "auc_roc": 0.97},
  intended_use="Consumer credit decisions under $100K",
  limitations="Not validated for business credit",
  input_schema={"income": "float", "debt_ratio": "float", "credit_score": "int"},
  output_schema={"decision": "enum[APPROVED, REJECTED]", "credit_limit": "float", "interest_rate": "float"}
)

Include model_hash (sha256) for artifact verification and schemas for downstream validation.