Running local models on Macs gets faster with Ollama’s MLX support - Ars Technica

LLMs Running local models on Macs gets faster with Ollama’s MLX support Apple Silicon Macs get a performance boost thanks to better unified memory usage. 64 A graphic made by Ollama to announce MLX support. Credit: Ollama A graphic made by Ollama to announce MLX support. Credit: Ollama Text settings Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only   Learn more Minimize to nav Ollama, a runtime system for operating large language models on a local computer, has introduced support for Apple’s open source MLX framework for machine learning. Additionally, Ollama says it has improved caching performance and now supports Nvidia’s NVFP4 format for model compression, making for much more efficient memory usage in certain models. Combined, these developments promise significantly improved performance on Macs with Apple Silicon chips (M1 or later)—and the timing couldn’t be better, as local models are starting to gain steam in ways they haven’t before outside researcher and hobbyist communities. The recent runaway success of OpenClaw—which raced its way to over 300,000 stars on GitHub, made headlines with experiments like Moltbook and became an obsession in China in particular—has many people experimenting with running models on their machines. As developers get frustrated with rate limits and the high cost of top-tier subscriptions to tools like Claude Code or ChatGPT Codex, experimentation with local coding models has heated up. (Ollama also expanded Visual Studio Code integration recently.) The new support is available in preview (in Ollama 0.19) and currently supports only one model—the 35 billion-parameter variant of Alibaba’s Qwen3.5. Hardware requirements are intense by normal users’ standards. Users need an Apple Silicon-equipped Mac, sure, but they also need at least 32GB of RAM, according to Ollama’s announcement. Further, Ollama now takes advantage of the new Neural Accelerators in Apple’s M5-series GPUs, so those brand-new Macs should see extra advantages in both tokens-per-second and time-to-token. Local models still lag behind frontier models in benchmarks, but we’re getting to the point that they’re good enough for some tasks users might normally pay a subscription for—and of course, there are privacy advantages to running models locally compared to cloud-based services, though we definitely do not recommend OpenClaw-like setups that give models deep access to your system. The main barriers remain setup (Ollama is first and foremost a command-line tool, though other interfaces have been made available) and hardware capabilities, especially video memory. Apple’s MLX offers optimized access to the memory on Apple’s chips, which is shared between the GPU and CPU—a different approach from the desktop machines with dedicated GPUs that Ollama has targeted before. This by no means closes the gap between cloud models and local ones for most users, but it’s potentially a step in the right direction for modern Mac users. Ollama hasn’t shared a timeline for when MLX support will exit preview and branch out to more models.