Model Version — AI Glossary

Reference details

Related terms

• Model Snapshot
• Model Family
• API Versioning

Example: Nothing in your code changed, and the behaviour did

A floating alias points at whatever the provider currently considers the latest release in a family; a dated identifier points at one frozen artifact. Ship against the alias and your evaluation suite validated a model that can be replaced underneath you — no deploy, no diff, no changelog entry in your repository, and a regression that bisects to nothing. Pinning the dated version makes the upgrade an explicit change you make, evaluate and can revert. The cost of pinning is that you must actually track deprecations, because snapshots are retired on a schedule and a pin that expires fails closed.

Commonly confused with

The model family is the product name, the alias is a moving pointer, and the version string is a frozen artifact. Only the third is reproducible. A model version is also not a version of *your* system — the prompt, the tools and the retrieval corpus all move independently, so pinning the model alone does not make a pipeline reproducible.

When to use it

Reach for it when:

• Production, always — pin the dated version and upgrade deliberately
• Recording which version produced a result, in evaluations and in traces
• Planning upgrades around published deprecation windows rather than discovering them

Reach for something else when:

• Prototyping, where the alias is convenient and the risk is nil
• Pinning and then never revisiting: retired snapshots stop working
• Assuming a pinned model makes the whole pipeline deterministic

Continue your research

Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Model Version — AI Glossary.

• Browse models
• Review providers
• Inspect benchmarks
• Explore workflows
• Learn AI concepts
• Read decision guides
• Review trust and methodology
• Start a decision path