RunAnywhere SDK v0.17.5: Cross-Platform On-Device AI

RunAnywhere released SDK v0.17.5, a cross-platform update that strengthens its on-device AI infrastructure across Swift (iOS/macOS), Kotlin (Android), Flutter, and React Native. The release unifies core runtime components and backend integrations, including llama.cpp for local LLM execution and ONNX-powered speech pipelines for voice activity detection, speech-to-text, and text-to-speech. The result is a more consistent, production-ready foundation for teams building chat, voice, and multimodal AI experiences directly on device.

Rather than focusing on incremental feature changes, v0.17.5 reinforces RunAnywhere's broader mission: make on-device AI the default way apps ship intelligence. By simplifying runtime alignment and backend packaging across platforms, this update reduces integration friction for mobile engineers, improves determinism for enterprise teams, and enables privacy-first, offline-capable AI experiences without the operational complexity typically associated with edge deployments.

What's New in v0.17.5

Unified Cross-Platform Runtime

This release brings consistency across all supported platforms:

Swift (iOS/macOS): Enhanced native integration with Core ML and Metal acceleration
Kotlin (Android): Optimized JNI bindings and improved memory management
Flutter: Streamlined plugin architecture for both iOS and Android
React Native: Updated native module bindings with better TypeScript support

Enhanced Backend Support

v0.17.5 strengthens the underlying inference engines:

llama.cpp integration: Updated to the latest version for improved LLM performance and broader model compatibility
ONNX Runtime: Optimized speech pipelines for VAD, STT, and TTS with reduced latency
Model packaging: Improved versioning and differential updates for over-the-air deployments

Production-Ready Infrastructure

The update includes several improvements for enterprise deployments:

More deterministic behavior across device fleets
Reduced memory footprint for constrained devices
Better thermal management and battery optimization
Enhanced error handling and recovery mechanisms

Why This Matters

On-device AI is no longer experimental. Teams are shipping production features that require:

Privacy by default: Keep sensitive data on-device without cloud round-trips
Offline capability: Work on planes, subways, and areas with poor connectivity
Low latency: Deliver instant responses that feel native, not cloud-dependent
Cost control: Reduce cloud inference bills by running models locally

v0.17.5 makes these capabilities more accessible by reducing the complexity traditionally associated with edge AI deployment. Instead of maintaining separate implementations for each platform, teams can now use a unified SDK with consistent APIs and behavior.

Getting Started

SDK v0.17.5 is available now across all platforms:

Swift (iOS/macOS): Quick Start Guide
Kotlin (Android): Quick Start Guide
Flutter: Quick Start Guide
React Native: Quick Start Guide
GitHub Release: v0.17.5 Release Notes

What's Next

This release sets the foundation for upcoming features:

Vision model support for multimodal experiences
Enhanced hybrid routing with fine-grained policy controls
Expanded model library with domain-specific optimizations
Advanced analytics and fleet monitoring capabilities

We're committed to making on-device AI the default choice for mobile and edge applications. v0.17.5 is another step toward that vision.

Feedback and Support

We'd love to hear how you're using the SDK:

GitHub Issues: Report bugs or request features
Discord Community: Join discussions with other developers
Email: Reach out to our team at support@runanywhere.ai

Download SDK v0.17.5 today and start building privacy-first, offline-capable AI experiences.

View the full release notes and download the SDK on GitHub: v0.17.5 Release