Ollama taps Apple’s MLX framework to make local AI models faster on Macs

Ollama's latest release integrates Apple's MLX framework to accelerate local LLM inference on Apple Silicon Macs, leveraging the shared CPU/GPU memory architecture to reduce latency and improve throughput. The update also adds support for NVIDIA's NVFP4 low-precision format, enabling larger models to run under tighter memory constraints. Currently MLX support is limited to the Qwen3.5-35B-A3B model, with more expected. The release is framed in the context of growing demand for local AI agents like OpenClaw, where running models locally offers data control and cost savings, though typically at slower speeds than remote APIs.