| Port details |
- colibri Run GLM-5.2 (744B MoE) on a 25GB-RAM consumer machine
- g20260717 misc
 =0      Package not present on quarterly.This port was created during this quarter. It will be in the next quarterly branch but not the current one. - Maintainer: yuri@FreeBSD.org
- Port Added: 2026-07-14 22:15:15
- Last Update: 2026-07-18 06:37:29
- Commit Hash: cf6b05e
- License: APACHE20
- WWW:
- https://github.com/JustVugg/colibri
- Description:
- colibri allows to run GLM-5.2 (744B-parameter MoE) on a consumer machine with
~25 GB of RAM - in pure C, with zero dependencies, by streaming experts from
disk.
Colibri is a lightweight, quality-preserving MoE runtime that treats VRAM, RAM,
and storage as one managed memory hierarchy. Insufficient fast memory may reduce
speed, but the default policy never silently changes model precision or router
semantics.
  ¦  ¦  ¦  ¦ 
- Manual pages:
- FreshPorts has no man page information for this port.
- pkg-plist: as obtained via:
make generate-plist - USE_RC_SUBR (Service Scripts)
-
- Dependency lines:
-
- colibri-ai>0:misc/colibri
- To install the port:
- cd /usr/ports/misc/colibri/ && make install clean
- To add the package, run one of these commands:
- pkg install misc/colibri
- pkg install colibri-ai
NOTE: If this package has multiple flavors (see below), then use one of them instead of the name specified above.- PKGNAME: colibri-ai
- Flavors: there is no flavor information for this port.
- distinfo:
- TIMESTAMP = 1784348501
SHA256 (JustVugg-colibri-g20260717-8a1dcca_GH0.tar.gz) = 78a4899981bd2da030b48efa328c3c6b330252b08425d154f58acfbbd23edf6a
SIZE (JustVugg-colibri-g20260717-8a1dcca_GH0.tar.gz) = 3055854
Packages (timestamps in pop-ups are UTC):
- Dependencies
- NOTE: FreshPorts displays only information on required and default dependencies. Optional dependencies are not covered.
- Build dependencies:
-
- gmake>=4.4.1 : devel/gmake
- python3.12 : lang/python312
- Test dependencies:
-
- python3.12 : lang/python312
- Runtime dependencies:
-
- py312-huggingface-hub>0 : misc/py-huggingface-hub@py312
- py312-safetensors>0 : misc/py-safetensors@py312
- py312-numpy>0 : math/py-numpy@py312
- py312-psutil>0 : sysutils/py-psutil@py312
- py312-pytorch>0 : misc/py-pytorch@py312
- py312-datasets>0 : misc/py-datasets@py312
- py312-tokenizers>0 : textproc/py-tokenizers@py312
- py312-transformers>0 : misc/py-transformers@py312
- python3.12 : lang/python312
- There are no ports dependent upon this port
Configuration Options:- No options to configure
- Options name:
- misc_colibri
- USES:
- gmake python shebangfix
- pkg-message:
- For install:
- You installed colibri: GLM-5.2 AI model runner.
It requires a very large download of ~400GB from HuggingFace:
https://huggingface.co/mateogrgic/GLM-5.2-colibri-int4-with-int8-mtp
Install py312-huggingface-hub and then:
$ hf download mateogrgic/GLM-5.2-colibri-int4-with-int8-mtp
- Master Sites:
|
Number of commits found: 4
| Commit History - (may be incomplete: for full details, see links to repositories near top of page) |
| Commit | Credits | Log message |
|---|
g20260717
18 Jul 2026 06:37:29
 |
Yuri Victorovich (yuri) |
misc/colibri: update g20260715 → g20260717
Also:
* add service
* upgrade optimization O2->O3
* add Python dependencies |
g20260715
16 Jul 2026 04:06:11
|
Yuri Victorovich (yuri) |
misc/colibri: update g20260714 → g20260715 |
g20260714
15 Jul 2026 08:12:59
|
Yuri Victorovich (yuri) |
misc/colibri: Minor change in the comment |
g20260714
14 Jul 2026 22:13:26
|
Yuri Victorovich (yuri) |
misc/colibri: New port: Run GLM-5.2 (744B MoE) on a 25GB-RAM consumer machine |
Number of commits found: 4
|
As an Amazon Associate I earn from qualifying purchases.