Burn is a Rust tensor library and deep learning framework for numerical computing, training and inference. It offers a unified API across CPU, CUDA, ROCm, Metal, Vulkan and WebGPU backends, with autodiff, kernel fusion, ONNX and PyTorch weight import, plus browser and no_std support.
Lemonade is a local AI server that runs optimized LLMs, speech, image and 3D models on your own GPU or NPU. It offers OpenAI, Anthropic and Ollama compatible APIs, a CLI, a model manager, and an embeddable binary for packaging local AI into apps.