Metal
Metal at Beyond Market Intelligence is a file of 3 stories. The newest of them: “Bring GPU-accelerated nearest-neighbor search to Apple Silicon with mlxVS”, “Google's ML Drift Opens On-Device AI to Any GPU Environment”, and “Cross-platform AI compression that stays consistent across any hardware”. If you're feeling constrained by running GPU-accelerated nearest-neighbor search only on NVIDIA hardware, mlxVS changes that for Apple Silicon users. Google has open-sourced ML Drift, a cross-platform GPU compute engine that finally lets on-device AI run on any GPU environment. Acme AI is the next-generation, AI-powered spreadsheet platform built to replace Excel and redefine how analysts, data scientists, and enterprise teams work… The list below is every Metal story on Beyond Market Intelligence, newest first.
Bring GPU-accelerated nearest-neighbor search to Apple Silicon with mlxVS
If you're feeling constrained by running GPU-accelerated nearest-neighbor search only on NVIDIA hardware, mlxVS changes that for Apple Silicon users. This open-source library ports NVIDIA cuVS algorithms, exact k-NN, k-means, NN-Descent graphs, and CAGRA, to MLX, delivering impressive speed on a MacBook Pro M3 Pro. It's a practical, accessible tool for anyone exploring faster vector search without leaving the Apple ecosystem. For deeper coverage on AI model behavior, see "Three Small AI Models Falter When Told Their Answers Are Wrong."
Google's ML Drift Opens On-Device AI to Any GPU Environment
Google has open-sourced ML Drift, a cross-platform GPU compute engine that finally lets on-device AI run on any GPU environment. It abstracts the messy complexity of OpenGL ES, OpenCL, Metal, and WebGPU behind a single Apache 2.0 license. This is the kind of practical infrastructure move that makes AI more accessible. For deeper context on how AI can optimize toolchains, see our coverage of "A 40% Speed Gain: How AI Optimized a Developer Toolkit Overnight."
Cross-platform AI compression that stays consistent across any hardware
One hash across seven backends proves that integer-only inference can solve the cross-platform compression problem that has long plagued LLM deployment. When float GPT-2 encoded on a Mac decoded zero of four windows on x86, the root cause was clear: floating-point inconsistency across hardware. This project's solution, int8 weights, fixed rounding, and lookup tables, delivers exact reproducibility without sacrificing performance. For deeper context on building efficient AI systems, our "Building a GPT-2 KV Cache" piece explores similar performance measurement challenges.