Signed in
We value your privacy
This site uses cookies to improve your browsing experience, analyze site traffic, and show personalized content.
| No holders yet. | ||||||||||
Colibrì is an experimental open-source inference engine designed to run extremely large Mixture-of-Experts (MoE) language models—such as GLM-5.2 (744B parameters)—on consumer hardware by streaming expert weights from disk instead of requiring the entire model in RAM or VRAM. 25k stars and stacked dev Redirected all fees to his github Github - https://github.com/JustVugg/colibri