JustVugg/colibri

A hierarchical inference engine for frontier MoE models: hot experts on the GPU (CUDA, Vulkan), warm ones in RAM, cold ones streamed from NVMe. Written in C, runs on the hardware you already own.

937 best points · 241 comments · 3 Hacker News posts · 48,543 GitHub stars · View on GitHub

Hacker News discussions

• Show HN: Getting GLM 5.2 running on my slow computer (937 points, 240 comments, by vforno, 2026-07-09)
• Call for probes: building the GLM-5.2 Expert Atlas (1 points, 0 comments, by Bluestein, 2026-08-11)
• Brio mode in Colibri: Scoring a closed set instead of generating (1 points, 1 comments, by vforno, 2026-09-21)