DeepSeek 4 Flash Local Inference Engine for Metal | Hacker News

A Hacker News thread focused on DeepSeek Flash running locally at high speed on Apple Silicon Macs.

Users shared measurements of power draw and token generation rates, comparing efficiency across consumer hardware.

The discussion weighed the economics and capability of local open-source models against hosted frontier models.

It also highlighted growing interest in practical on-device AI and LLM optimization on Macs and other personal machines.