# JustVugg/colibri

> A pure-C, zero-dependency inference engine for running large MoE models (744B–2.8T parameters) on consumer hardware by treating VRAM, RAM, and storage as a single multitier hierarchy and streaming experts from disk.

- Magnitude: 4.3 out of 10 — Steady
- Stars: 44,732 total · +37 stars measured 2026-10-29, 00:03–02:53 UTC, ≈ 303 by evening
- Star trust: star growth looks organic
- Category: Language models · Language: C · License: Apache-2.0 · Created: 2026-07-01 · Last push: 2026-10-29
- GitHub: https://github.com/JustVugg/colibri · Homepage: https://justvugg.github.io/colibri · Page: https://gitnova.dev/en/r/JustVugg/colibri

## Useful for

- Run a chat with the 744B GLM-5.2 model locally via coli chat
- Launch the coli web dashboard to watch expert routing and token metrics
- Experiment with placing experts across VRAM, RAM, and NVMe to speed up decoding

## Why it’s here

- Star-counter measurements on 2026-10-29 (UTC), 00:03–02:53: 44695 → 44732 stars (+37). This is the change over that interval.
- Estimated end-of-day forecast: about +303 stars, using observed gains and the previous day.
- Over the last two days the pace is 1.7× that of the previous week and a half.
- About 62 forks a day — people are taking the code.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 5,087
- Issues and pull requests: 2,380
- Watchers: 355
- Average over the last week: 234 per day
- Usual pace: 219 per day
- Stars in the last hour (measured): 16
- Latest release: v2.0.0 (2026-10-06)

## Stars per day, last 30 days (oldest → newest, today is partial)

2026-09-30 … 2026-10-29: 178, 205, 252, 234, 241, 247, 277, 291, 237, 245, 218, 250, 272, 221, 182, 281, 221, 179, 148, 157, 177, 157, 204, 204, 122, 185, 169, 343, 308, 37

## Magnitude by day, last 14 days

10-16 3.9, 10-17 3.5, 10-18 3.5, 10-19 3.4, 10-20 3.4, 10-21 3.5, 10-22 3.7, 10-23 4.0, 10-24 3.9, 10-25 3.4, 10-26 3.5, 10-27 3.6, 10-28 4.6, 10-29 4.3

## Similar by description

1. **mw00/project-maya** — 2.7 · Steady · Language models · Cuda · +7 stars measured 2026-10-29, 00:05–02:55 UTC, ≈ 58 by evening
   Local engine, server and dashboard to run the 321B MoE model GLM-5.3-Flash on one or two NVIDIA GPUs, offloading experts to RAM and NVMe.
   Full card: https://gitnova.dev/en/r/mw00/project-maya.md
2. **FlashML-org/FreeToken** — 1.8 · Steady · Language models · Python · +1 star measured 2026-10-29, 00:09–02:59 UTC, ≈ 15 by evening
   A Mixture-of-Experts inference engine for running frontier-scale open-weight MoE models (290B+) on consumer hardware — GPUs, CPUs, and host memory. Aimed at users who want to run frontier models locally.
   Full card: https://gitnova.dev/en/r/FlashML-org/FreeToken.md
3. **prabanta-dev/janas** — 0.2 · Steady · Language models · C · +0 stars measured 2026-10-29, 00:12–03:03 UTC, ≈ 0 by evening
   Janas is a family of AI tools in C with no dependencies. Janas-LLM runs large MoE models (up to 80B, 48 GB) on a 32 GB laptop by streaming experts from NVMe SSD, and outperforms llama.cpp on all 13 tested models.
   Full card: https://gitnova.dev/en/r/prabanta-dev/janas.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-10-29 03:04 UTC, updated every 30 minutes.