Model Management and Inference
Model Versioning — References
Model versioning › References
Sources on version identifiers, deprecation policy, and artefact immutability.
Provider deprecation and snapshot policy
- OpenAI deprecations — the published list of retired and scheduled-for-retirement model identifiers with dates.
- Anthropic model deprecations — retirement schedule and the distinction between dated snapshots and aliases.
- Google Gemini API documentation — model identifiers, stable versus preview designations, and version lifecycle notes.
Artefact identity
- Hugging Face Hub documentation — repository revisions, model cards, and pinning a model to a commit rather than a branch.
- OCI Image Specification — content-addressed, digest-pinned artefacts, the model most self-hosted weight distribution converges on.
- Semantic Versioning 2.0.0 — a useful contrast rather than a template: model versions carry no compatibility guarantee, so a semver-shaped label on a model is a claim the artefact cannot support.
Lineage tooling
- MLflow Model Registry documentation — stages, versions, and lineage for registered models. Cited by name: the documentation site reorganises its paths between releases.
Provider deprecation pages change structure; treat the date in a retirement notice as the authority and re-read the page rather than relying on a cached summary.