Colibri
Pure C inference engine for running mixture of experts models on any hardware.
0 members list Colibri, unchanged over the last 12 weeks, data as of 21 November 2026
Members who list Colibri in their Stack, by week.
Members who list Colibri
Be the first to add Colibri
Nobody on Stackness runs Colibri yet. Add it to your stack and you're the first member running it.
About Colibri
Colibri is a C-based inference engine designed to run mixture of experts language models on standard hardware without requiring specialized equipment. It streams model weights from disk rather than loading them entirely into memory, enabling large models to run on machines with limited resources. The tool supports multiple operating systems including Windows, Linux, and macOS, and can optionally leverage GPUs when beneficial for performance. Developers use it to execute open-source models locally while retaining control over their hardware and data.
Colibri is developed by JustVugg and released under the Apache 2.0 license. The engine is written in pure C with no external dependencies, making it portable and straightforward to deploy. It includes a setup process that analyzes available system resources and recommends suitable models, handles model downloads with resume capability, and provides both a dashboard interface and API compatibility with OpenAI and Anthropic standards.
Checked 16 November 2026.
Nobody on Stackness runs Colibri yet.
Sign up and add itPosts about Colibri
- GLM 5.3 Flash as a coding agent: a month of daily use, $68, and the slot it kept
A Wagtail maintainer ran 1B tokens of GLM 5.3 Flash in September for $68 and kept it as the default for 95% of tasks. The costs, the setup and what breaks.