Introduction to Mirai

Mirai is the unified inference layer for the intelligent edge. It enables developers to deploy and run state-of-the-art open source models directly on user devices with a single API, handling hardware acceleration (NPU/GPU), quantization, and model caching automatically.

Edge First

Requests are routed to the local device NPU by default. This ensures zero latency, zero marginal cost, and full offline capability.

Cloud Fallback

Seamlessly routes complex reasoning tasks or requests from underpowered devices to our optimized server-side GPU clusters.

Mirai is compatible with iOS 16+, Android 12+, and modern browsers supporting WebGPU.