Today we could run this beast of 1 trillion tokens, Kimi 2.5 thanks to : https://huggingface.co/DevQuasar/moonshotai.Kimi-K2.5-GGUF for the IQ2_XXS 267 GB version.

It runs ~1 token/s with a 256gb ram and some "paging memory" (using hard disk as ram).

Currently under test, we are very excited to see the results. It is quite amazing to be able to run All available models with this build of HugstonOne Enterprise Edition 1.0.8 but most importantly, Deepseek 3.1 Terminus, Qwen 3 80b and 235b and now Kimi 2.5.

Write your questions if any, otherwise enjoy.

Hugston Team.