PyTorch has released the ExecuTorch MLX Delegate, an experimental backend that enables GPU-accelerated inference for PyTorch models on Apple Silicon Macs using Apple's MLX framework. It integrates with the PyTorch 2 export stack via torch.export and supports quantization options including BF16, FP16, FP32, and 2/4/8-bit affine quantization via TorchAO. The delegate delivers 3-6x higher throughput compared to existing ExecuTorch backends on macOS. Validated models include dense transformers (Llama 3.2, Qwen 3, Gemma 3, Phi-4 mini), sparse Mixture-of-Experts (Qwen 3.5 35B-A3B), and speech-to-text models (Whisper, Parakeet, Voxtral) for both offline and real-time transcription. The workflow follows the standard ExecuTorch pipeline: export with torch.export, lower with MLXPartitioner, and run the resulting .pte file.
