colibri
colibri: Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦. It is ranked #132 on the Stars Radar, in Models & inference. Agentik OS starred it 50 days ago, and it has 48.3k GitHub stars.
What colibri says about itself
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.
Alternatives to colibri
- heretic — Fully automatic censorship removal for language models
- Edge0 — Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac,…
- litellm — The fastest, litest AI Gateway. Rust core with Python SDK. Call 100+ LLM APIs in OpenAI (or native) format with cost…
- Sana — SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformer
- BitNet — Official inference framework for 1-bit LLMs
- PowerInfer — High-speed Large Language Model Serving for Local Deployment
colibri in numbers
- Rank on the Stars Radar: #132 of 314
- 48.3k GitHub stars
- Starred by Agentik OS 50 days ago, last push 57 min ago
- Owner: JustVugg on GitHub
- Market: Models & inference
FAQ
What is colibri?
Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦 Agentik OS starred it on GitHub 50 days ago; it is ranked #132 on the Stars Radar.
Is colibri free?
It is an open-source repository on GitHub; check its license before use.
Who makes colibri?
The GitHub account JustVugg. The repository has 48.3k GitHub stars.
Every repository Agentik OS stars on GitHub is read and classified by AGK Intelligence: what it is, which market it belongs to, how useful and how mature. 314 repositories, 10 markets. The stars are read every 5 minutes — last update 2026-11-06 13:35 UTC. Full method.