# What changed in @ai-sdk/mistral

> 17 changes in @ai-sdk/mistral 4 from its official release notes, upgrade guides and advisories: 1 security, 4 new APIs and options. Each one cited. Sources checked 2026-10-18. HTML: https://unstale.sh/npm/@ai-sdk/mistral/

To see which of these apply to a repository, run `npx @unstale/cli init` in it.

## @ai-sdk/mistral 4

### Security

- @ai-sdk/provider-utils has an Uncontrolled Resource Consumption issue (4.0.33) — [security advisory](https://github.com/advisories/GHSA-866G-F22W-33X8)

### New APIs and options

- Adds a conversations API to @ai-sdk/mistral. (4.0.60) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.60)
- Adds a web search tool to @ai-sdk/mistral. (4.0.60) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.60)
- The portable reasoning levels now include max, widening LanguageModelV4CallOptions['reasoning'] and ReasoningLevel. Consumers with exhaustive switches or mappings must handle it. (4.0.58) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.58)
- Experimental streaming lifecycle callbacks were added for speech generation and transcription. (4.0.56) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.56)

### Behavior and default changes

- Top-level embedding dimensions are mapped to provider settings. (4.0.62) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.62)
- Telemetry was added to speech generation and transcription, with provider usage propagation. (4.0.56) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.56)

### Fixes and other changes

- Reasoning configuration is forwarded for unknown model IDs instead of being silently stripped; known models without reasoning-effort support retain their previous handling. (4.0.63) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.63)
- Reasoning effort is now enabled for Mistral Large 4 models. (4.0.59) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.59)
- JSON schemas are preserved when structured outputs are disabled. (4.0.55) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.55)
- Provider error finish reasons are mapped to error. (4.0.55) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.55)
- The User-Agent header is now standards-compliant. (4.0.54) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.54)
- Fixes streamed tool calls with missing, blank, or repeated IDs. (4.0.53) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.53)
- Updated Mistral model IDs. (4.0.43) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.43)
- Updated the Mistral reasoning configuration. (4.0.43) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.43)
- Raw usage metadata now preserves complete Mistral chat usage objects. (3.0.63, 4.0.36) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%403.0.63), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.36)
- Mistral now supports incremental streaming tool calls. (2.0.43, 3.0.59, 4.0.31) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%404.0.31), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%403.0.59), [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/mistral%402.0.43)