Release 0.13.0

0.13.0 (2026-11-11)

This release moves every provider to AI SDK 7, adds the iOS 27 Apple Intelligence APIs, and ships three new packages. It also adds a CHANGELOG.md to each package. Read those for the full list of changes.

New packages

• @react-native-ai/core-ai runs Apple Core AI model bundles on iOS 27. It supports language, vision, image and video generation, transcription, embeddings, segmentation and raw model calls.
• @react-native-ai/android runs Gemini Nano on ML Kit GenAI. It supports generation, streaming, images, tool calling, transcription and embeddings.
• @react-native-ai/utils adds wrapLanguageModelWithHistory, which keeps long conversations within any model's context.

Highlights

• @react-native-ai/apple adds:
  • Private Cloud Compute, image prompts, Image Playground and Vision tools
  • reasoning levels and token usage
  • required tool choice and live transcription
  • an Expo config plugin, plus many fixes from on-device testing
• @react-native-ai/llama adds reasoning, structured output, multimodal input and countTokens, along with crash fixes found on device.
• @react-native-ai/dev-tools works with AI SDK 7 telemetry.

Breaking changes

• Every provider requires AI SDK 7 (ai@7). Don't upgrade if you are still on AI SDK 6.
• @react-native-ai/llama requires llama.rn 0.13.0-rc.8 or later.
• In @react-native-ai/apple, isAvailable() is removed in favor of getModelInfo(), and AppleEmbeddings.getInfo() is now getModelInfo().
• For dev tools, register the tracer once with registerTelemetry(...) instead of passing experimental_telemetry: { tracer } on each call.

Removed

• @react-native-ai/adk is replaced by @react-native-ai/android.
• @react-native-ai/mlc is no longer maintained. Use @react-native-ai/llama instead.