RunAnywhere SDKs v0.20.9

Cross-platform on-device AI SDKs — iOS/macOS (Swift), Android (Kotlin), Flutter, React Native, and Web — plus the rcli desktop CLI. A single C/C++ core (runanywhere-commons) behind a stable C ABI; each SDK is a thin platform bridge.

Backends (all platforms)

• llamacpp — LLM + VLM (GGUF)
• sherpa — STT / TTS / VAD (ONNX)
• onnx — embeddings
• Apple MLX — source-built in the Swift SDK (Apple LLM/VLM/embed/STT/TTS)

Availability

• Swift (SPM): resolve this repo at v0.20.9 — remote XCFrameworks, checksum-verified.
• rcli: macOS-arm64 + Linux-x86_64 tarballs attached below.
• Kotlin (Maven Central): io.github.sanchitmonga22:runanywhere-{sdk,llamacpp,onnx}-android:0.20.9
• npm: @runanywhere/{web,web-llamacpp,web-onnx,core,llamacpp,onnx,proto-ts}@0.20.9
• pub.dev: runanywhere{,_llamacpp,_onnx}: ^0.20.9

Notes

• The prebuilt Flutter/RN MLX binary packages are held for a 0.20.10 follow-up (build-reproducibility work on RunAnywhereMLXRuntime). Swift SPM MLX is unaffected (source-built).
• Windows native is advisory/experimental for this release.
• QHexRT (Qualcomm Hexagon NPU runtime) ships separately as a private release, version-aligned at v0.20.9.

Published 2026-07-13.