# Mistral Small 3.2 — deprecation status

> Yes — Mistral Small 3.2 is retired. Mistral deprecated it on April 30, 2026 and shut it down on July 31, 2026; API requests to it now fail. The recommended replacement is Mistral Small 4.

The last of the Mistral Small 3 line, a June 2025 point release that mostly fixed instruction-following and repetition problems in 3.1 rather than adding anything. It outlived its three predecessors by eight months and stopped answering on 31 July 2026, which closed out the 3.x Small series entirely.

## Facts

| Field | Value |
| --- | --- |
| Provider | Mistral |
| Model id | `mistral-small-2506` |
| Status | Retired |
| Deprecated | April 30, 2026 |
| Shutdown | July 31, 2026 (3 days ago) |
| Last verified | August 3, 2026 |

## Replacement

- **Mistral Small 4** (`mistral-small-4-0-26-03`) · recommended — https://modeldeprecations.dev/mistral/mistral-small-4-0-26-03
  - Small 4 is a hybrid reasoning model. It can emit thinking content that Small 3.2 never produced, so parsers that assumed the whole response body was the answer need checking.
  - Parameters it accepts: https://modelparams.dev/models/mistral/mistral-small-4-0-26-03

```diff
- "model": "mistral-small-2506"
+ "model": "mistral-small-4-0-26-03"
```

## FAQ

### When did Mistral Small 3.2 shut down?

Mistral Small 3.2 shut down on July 31, 2026. API requests to it fail from that date onward.

### What replaces Mistral Small 3.2?

Mistral recommends Mistral Small 4 (`mistral-small-4-0-26-03`) as the replacement for Mistral Small 3.2. Small 4 is a hybrid reasoning model. It can emit thinking content that Small 3.2 never produced, so parsers that assumed the whole response body was the answer need checking.

### Is Mistral Small 3.2 still available?

No. Mistral Small 3.2 is retired and no longer served — it was shut down on July 31, 2026. Calls to it return an error.

### How do I migrate from Mistral Small 3.2 to mistral-small-4-0-26-03?

Swap the model id in your API call: replace "mistral-small-2506" with "mistral-small-4-0-26-03". Small 4 is a hybrid reasoning model. It can emit thinking content that Small 3.2 never produced, so parsers that assumed the whole response body was the answer need checking.

## Sources

- [Mistral models overview](https://docs.mistral.ai/getting-started/models/models_overview/) — accessed 2026-08-03

---

Canonical page: https://modeldeprecations.dev/mistral/mistral-small-2506
JSON: https://modeldeprecations.dev/api/v1/models/mistral/mistral-small-2506.json