Running DeepSeek 4.1 Flash on a 32GB Mac via SSD streaming with open-source Local Code

alexcovo_eth · x · 2026-11-14

Developer ivanfioravanti demos DeepSeek 4.1 Flash running on a 32GB Mac using Local Code's SSD streaming: model weights (376GB on disk) stay on the SSD and load on demand instead of fitting fully in RAM (64GB recommended, 32GB minimum).

Local Code is an open-source, fully local Mac AI agent app: nothing leaves the device, it works offline, cloud models are off by default, and the agent runs Python/Node.js/shell in an Apple sandbox with access only to folders you share. It's tuned for Apple Silicon with per-memory-tier model recommendations.

Original post →

More from coding & agent

coding & agent channel →