Nexa SDK

Nexa SDK is an on-device AI inference framework that lets you run any model on any device — completely offline, with no cloud dependency. Whether you're working on a CPU, GPU, or NPU, Nexa SDK has you covered with backend support for CUDA, Metal, Vulkan, and Qualcomm NPU. It handles multiple input modalities including text, images, and audio, making it a genuinely versatile foundation for local AI development. The SDK comes with an OpenAI-compatible API server, supports JSON schema-based function calling and streaming, and is compatible with model formats including GGUF, MLX, and Nexa AI's own .nexa format for efficient quantized inference. If you're building AI applications that need to run fast, private, and on-device, Nexa SDK is the infrastructure layer you've been looking for.