proprietary

Definition

Closed models and data: available only through an API, with weights and training corpora undisclosed. The label matters for reproducibility — outsiders can measure behaviour but not inspect or fix it — and it defines the target in work that tries to copy such systems from their outputs alone. The opposite pole is open-weight; this is about disclosure, not price.

Explainers using this term

• Evaluating Distilled Models — Is "Close to the Teacher" a Good Metric?The False Promise of Imitating Proprietary LLMs