A quantized ONNX version of the multilingual MiniLM paraphrase embedding model, optimized for efficient deployment.