# colibrì

> Apache-2.0 pure-C zero-deps MoE streamed from disk

**Category:** AI Agents  
**Pricing:** open-source  
**URL:** https://justvugg.github.io/colibri/  
**GitHub:** https://github.com/JustVugg/colibri  
**Added:** 2026-09-16

---

## The Hook

colibrì runs frontier MoE models on hardware you already own — pure C engine, zero runtime deps, experts streamed from NVMe instead of living entirely in RAM/VRAM.

## What's Good

Apache-2.0 (JustVugg/colibri). Single-file C engine with disk/RAM/VRAM multitiering for huge MoE families (GLM-class and cousins). `coli chat` / `coli serve` / `coli web`. No BLAS, no Python at engine runtime. Site at justvugg.github.io/colibri. Prebuilt releases for Linux/macOS/Windows.

## The Catch

Vs Ollama/vLLM (local serve / high-throughput GPU serving): colibrì is research-grade MoE disk-streaming for models that do not fit, not the everyday pull-and-chat DX of Ollama or the production serving stack of vLLM. research/HW-heavy — expect fast NVMe, large model containers (hundreds of GB), and tokens/sec that track your disk hierarchy.

## Verdict

Apache-2.0 pure-C MoE engine. Disk-streamed experts vs Ollama/vLLM; hardware-honest.