# What changed in @ai-sdk/typesafe-ai

> 5 changes in @ai-sdk/typesafe-ai 3 from its official release notes, upgrade guides and advisories: 2 new APIs and options. Each one cited. Sources checked 2026-10-21. HTML: https://unstale.sh/npm/@ai-sdk/typesafe-ai/

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

## @ai-sdk/typesafe-ai 3

### New APIs and options

- Experimental decision calls accept maxOutputTokens and forward it to language-model-backed decisions. (3.0.23) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/typesafe-ai%403.0.23)
- Experimental decision calls accept reasoning and forward it to language-model-backed decisions. (3.0.23) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/typesafe-ai%403.0.23)

### Performance

- Construction is deferred until first use to reduce package import time. (3.0.24) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/typesafe-ai%403.0.24)

### Behavior and default changes

- Native decision providers warn when maxOutputTokens or reasoning is unsupported. (3.0.23) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/typesafe-ai%403.0.23)

### Fixes and other changes

- Language-model adapters keep reasoning disabled when the setting is omitted. (3.0.23) — [release notes](https://github.com/vercel/ai/releases/tag/%40ai-sdk/typesafe-ai%403.0.23)