AI · Technology · 2026-09-11 11:49:00
A developer posts a Show HN project, HungryGPU, which tracks local AI models, patches, and recipes by hardware. The tool assesses which merged pull requests, open-weight releases, community setups, or papers are actually valuable for a given hardware profile—and which are quietly left behind. Data is rebuilt every few hours from primary sources, with every tag carrying the literal line of evidence justifying it. The catalogue works without JavaScript, with one Atom feed per machine. It currently tags 1,225 merged pull requests, 1,052 open-weight checkpoints, 1,926 community setups, and 10,678 papers. The tool helps local AI users pick optimal models and configurations for their hardware.
Auto-generated summary for reference only. RSS