apple-neural-engine

Local Laya typed decisions on Apple Core ML and Neural Engine. Validated ports, ~5 ms short decisions on M3 Max, reproducible speed and energy benchmarks.

Run LLMs on Apple devices with CoreML, optimized for Apple Neural Engine + GPU

Your models on any xPU

PyTorch → CoreML conversion pipeline for Kokoro TTS. Unlocks fast on-device text-to-speech on Apple Neural Engine.

Python runtime for the Apple Neural Engine

ModernBERT model optimized for Apple Neural Engine.

Correctness-validated heterogeneous Laya runtime for Apple Silicon (MLX GPU + Apple Neural Engine)

Push-to-talk voice dictation for macOS. 100% local, free, open source. Apple Silicon MLX. No cloud, no subscription.

First super-resolution model designed for Apple Neural Engine. 2x upscale, real-time, on-device. Built by Ben Racicot.

A toolkit for BLE digital stethoscopes — capture, DSP, log-mel spectrograms, and murmur classification on the Apple Neural Engine

Apple Neural Engine Microarchitecture Study: M6 — report, raw data, tools and figures | Apple M6 神经网络引擎微体系结构研究：报告、原始数据、工具与插图

Real-time aerial person & vehicle detection, tracking & following for drones — custom YOLOv8 (ONNX + CoreML/ANE), occlusion re-ID, speed & size, ArduPilot.

Apple Neural Engine (ANE/NPU) vs GPU for local LLM inference on Apple Silicon / macOS. Core ML (CoreML), Core AI (CoreAI), MLX and Metal benchmarks: prefill, throughput, latency, memory, thermals, INT4/INT8, W4A16/A8W4 quantization, grouped scales and FP16 arithmetic. Reproducible component tests, compatibility findings, English/Chinese articles.

Windows port of Laya typed-decision AI (ONNX Runtime + DirectML) — Core ML/Apple Neural Engine alternative with first-class Arabic support. No text generation, no hallucination, runs on any DX12 GPU.

Meta's EdgeTAM (SAM 2 for the edge) as four Core ML packages: click once, track the object through video at ~48 ms a frame on the Apple Neural Engine.

Neural network that generates structured design specifications from type embeddings using ANE-accelerated training

ANE-optimized CoreML embeddings, converted and quantized from Hugging Face, maintained as a Swift package

First speculative decoding using Apple Neural Engine as draft + GPU as verifier on Apple Silicon. Custom minGRU encoder for heterogeneous compute.

Run text embedding and reranking models on the Apple Neural Engine of Apple Silicon Macs. OpenAI-compatible server; ModernBERT, BERT, and XLM-RoBERTa.

Run LLMs on Apple devices with CoreML, optimized for Apple Neural Engine + GPU