# Model Snapshot

> One fixed version of a hosted AI model, called by its own ID.

**Field:** Artificial Intelligence  
**Category:** LLMs  
**Also known as:** Pinned Model Version, Dated Model ID, Model Alias (LLM)

## Think of it like

It is like a photo of a cake on one day. The baker makes new cakes, but the photo never changes.

## Definition

A model snapshot is one fixed version of a hosted AI model, named by its own ID. The model behind that ID does not change, so the same prompt meets the same model next month. Some providers also offer an alias, a short name that moves to the newest snapshot. An alias is handy for trying things out, but it can change under you. Apps in real use usually name the snapshot, and move to a new one only after testing it.

## Example

On Anthropic's API, the alias claude-sonnet-4-5 points to claude-sonnet-4-5-20250929. A team that wants no surprises puts the dated name in its code.

## Example code

```json
{
  "model": "claude-sonnet-4-5-20250929",
  "max_tokens": 1024,
  "messages": [{ "role": "user", "content": "Hello" }]
}
```

This request names a dated snapshot, so the model behind it will not change.

## Why it matters

A new version can answer the same prompt in a new way. Naming a snapshot keeps your app steady until you choose to move.

## Common mistake

Using a moving alias in a live app. When it jumps to a new snapshot, answers can change with no warning.

## Three principles

- **One fixed model.** Same ID, same model.
- **An alias moves.** It follows the newest snapshot.
- **Each has an end.** Every snapshot retires one day.

## Related terms

- [Large Language Model (LLM)](https://learn.benibix.com/term/ai/large-language-model)
- [LLM Non-Determinism](https://learn.benibix.com/term/ai/llm-non-determinism)
- [Frontier Model](https://learn.benibix.com/term/ai/frontier-model)
- [Deprecation](https://learn.benibix.com/term/development/deprecation)
- [Semantic Versioning (SemVer)](https://learn.benibix.com/term/development/semantic-versioning)

## Taught in

- [Open-Weight vs Closed Models](https://learn.benibix.com/learn/ai-llms/open-weight-vs-closed-models): LLMs, 11/15
- [AI Model Names and Sizes](https://learn.benibix.com/learn/ai-models-providers/ai-model-names-and-sizes): AI Models and Providers, 6/13

![Three dated snapshots of one model sit in a row, oldest to newest. A short alias name has an arrow that points to the newest one, while the app's code points straight at one dated snapshot.](https://learn.benibix.com/diagrams/ai9-model-snapshot.svg)

---

Source: https://learn.benibix.com/term/ai/model-snapshot
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
