Release 0.12.0

Welcome in 2026 with a new release 👋

New features

• All providers are now V3‑ready for AI SDK v6
• Added Rozenite AI SDK profiler plugin
• Added tool calling support for Llama (streaming + non‑streaming)
• Added re-ranking support for llama models (new feature in v6)

Breaking changes

Provider APIs now target AI SDK v6 (V3 provider spec). Do not upgrade if you’re still on a lower SDK version.

Small breaking change in the Apple interface (should be low impact for most users). When working with embeddings, you must specify language (optionally) when getting an instance of a model.

const model = apple.textEmbeddingModel({ language: 'en' })

Previously, you would pass it alongside each call as providerMetadata.

providerMetadata: { apple: { language: 'en' } }

The following API was removed:

apple.textEmbeddingModel('NameOfSelectedModel')

It allowed choosing a different model by string, but only one model exists, so this was likely never used, hence we decided to introduce such breaking change to avoid introducing unnecessary legacy early in the project.

Release 0.11.0

0.11.0 (2025-10-14)

This release refactors MLC for Android (#134) (f145812), bringing precompiled binaries and much better (and feature complete) interface.

npm install @react-native-ai/mlc

MLC on Android is still experimental, with a few features missing. We encourage you to try it, using one of the built-in models to measure performance and return with feedback!

Release 0.10.0

0.10.0 (2025-09-27)

Features

• add support for Android in MLC script - this is preparation for Android support (landing soon) (#130) (44b2137)
• update to Expo SDK54 and drop beta from AI SDK (#132) (ccda8af)

Release 0.9.0

0.9.0 (2025-09-13)

Features

• Add support for tool calling (#120) (c4ba089)
• Implement much better and cleaner underlying code to handle streaming ee136ed)

Release 0.8.0

0.8.0 (2025-09-12)

Features

• add expo config plugin (#117) (b9c1141)
• add generation options support to MLC engine (#116) (ce7c4c3)
• structured output for MLC on iOS (#119) (dde9da6)

Release 0.7.1

0.7.1 (2025-09-11)

MLC on iOS has been refactored to provide better API and smoother experience.

Highlights include:

• AI SDK v5
• Precompiled runtime for most popular models, just install the package
• Expo config plugin
• New documentation with richer API coverage

Try it out and let us know what you think! We're open to your feedback!

---

Full Changelog: v0.6.0...v0.7.1

Release 0.6.0

0.6.0 (2025-09-07)

Bug Fixes

• example: add curent time tool (64341f9)

Features

• add voice selection (#108) (22a1fec)
• bottom tabs and ability to play audio buffers (#100) (c6a9ab0)
• improve example application and prepare for further features (#109) (3dd4ab9)
• new better demo app (#96) (4790476)
• rework example UI (#107) (ef9a805)
• working audio recorder (#103) (f05b9f1)

v0.5.0

TTS on iOS is here

What's Changed

• feat: separate speech synthesis from transcription and introduce new speech API by @grabbou in #93
• feat: text to speech by @grabbou in #95

Full Changelog: v0.4.0...v0.5.0

v0.4.0

Transcribe audio on iOS

You can now perform on-device transcription with built-in Apple models.

What's Changed

• docs: clarify React Native CLI example label by @shlee8405 in #87
• fix(mlc): release script by @szymonrybczak in #88
• feat(apple): transcribe by @grabbou in #89
• chore: fix TS setup by @jbroma in #92

New Contributors

• @shlee8405 made their first contribution in #87
• @jbroma made their first contribution in #90

Full Changelog: v0.2.0...v0.4.0

v0.3.0

Embeddings support on iOS

You can now generate embeddings with on-device models on iOS.

Full Changelog: v0.2.0...v0.3.0