TESIGN
SAVED

EN · KO

  1. TESIGN / RADAR
  2. 2026.09.13
  3. colibri

colibri

Run huge MoE models on your own PC

WHO USES ITML engineer · Infra owner · Systems developer

Install to use

Runs frontier MoE models on hardware you already own.

★ 33.8k▲ +974 / 7d▲ 937 HNDaily star increases over 14 days. Hollow bars mean no data. The last day may be partial.2026-09-03 no data2026-09-04 +72026-09-05 +72026-09-06 +332026-09-07 no data2026-09-08 no data2026-09-09 no data2026-09-10 +232026-09-11 +632026-09-12 +2162026-09-13 +1932026-09-14 +672026-09-15 +2612026-09-16 +151largest one-day increase in the 14 days+261last 14 days · stars per dayhollow = no data

33,753 stars · checked on GitHub GitHub check 30 h late · incl. +211 observed via GH Archive since then · 7-day +974 observed via GH Archive (as of ) · 937 Show HN points observed

RANKS All-time #185 Rising this week #7 INFRA #10

CROSS-SIGNAL Show HN + GitHubRISING +175 ★ / 24h

colibrì web dashboard: hardware and VRAM/RAM/disk tiers at left, a chat with a 744B model in the centre, tok/s metrics on top.
colibrì web dashboard: hardware and VRAM/RAM/disk tiers at left, a chat with a 744B model in the centre, tok/s metrics on top.
colibrì web dashboard: hardware and VRAM/RAM/disk tiers at left, a chat with a 744B model in the centre, tok/s metrics on top.
colibrì web dashboard: hardware and VRAM/RAM/disk tiers at left, a chat with a 744B model in the centre, tok/s metrics on top.

TESIGN TAKE

Lets you test on your own hardware whether placing a huge model across disk, RAM and VRAM actually works.

RUN

AT A GLANCE

LICENSE
Apache-2.0
USAGE
Commercial use, redistribution OK. Keep notices; mark changes.
LANGUAGE
C
PLATFORM
linux · macos · windows · cli
ACTIVITY
last commit 1 day ago () · latest release v1.11.0 () · 17 releases
COMMUNITY
150 contributors · 117 open issues (incl. PRs) · made by: an individual
SOURCES
Show HN + GitHub
OPEN SOURCE
YES
FIRST SEEN
CATEGORY
AI · INFRA

A summary, not legal advice.

WHY IT MATTERS

Large MoE models rarely fit in memory, and the usual answer is to buy different hardware. colibri streams expert weights from disk as they are needed, so the model runs on hardware you already own. It is written in pure C with no dependencies and published under Apache-2.0. Which models are supported has to be checked in the repository itself.

BUILD FROM THIS

  • An in-house local-model test bench on a single server
  • An inference binary built with no dependencies and shipped in a product
  • A comparison run before deciding on new hardware

WHO IT'S FOR

ML engineer
measures local inference limits before buying hardware
Infra owner
trials a large model on one in-house server
Systems developer
wants pure-C inference code to read and change

START IN 5 MINUTES

# Get a Linux/macOS/Windows build from Releases (Python 3 needed) or build:
$ git clone https://github.com/JustVugg/colibri && cd colibri/c
$ ./setup.sh
# Download a model (GLM-5.2 int4, about 372 GB) from Hugging Face, then:
$ COLI_MODEL=/nvme/glm52_i4 ./coli chat
$ ./coli web  --model /nvme/glm52_i4

CAVEATS

  • The reference model GLM-5.2 needs 372 GB of disk and 16 GB+ of RAM
  • Speed depends on the disk; a slow drive can mean under one token per second
  • A research engine that explicitly promises no SLA on speed

RECEIPT

FIRST SEEN
AT SOURCE
KEPT
STAR DELTA
+175 ★ / 24h
CREATED → FIRST SEEN
73d
WHEN FIRST SEEN
28,753 ★
SOURCES
Show HN + GitHub

The same facts in machine-readable form — View as Markdown · JSON

SIMILAR TOOLS

Up to five from AI by ★ total: edited entries first, then repository cards; this entry is highlighted. Drawn from the same stored snapshot as the rankings — a comparison, not a recommendation.

IMAGENAME★ TOTAL▲ 7dLICENSEPLATFORMLAST PUSH
ECC README banner: headline 'The operating system for AI agent harnesses', harness chips, and skill, agent and command lists.
eccSkills, memory and security for agents★ 259k▲ +832MITpermissivewindows · macos · linux · cli
Hermes Agent landing: blue background, the headline 'The agent that grows with you' and an engraving-style winged figure.
Hermes AgentSelf-improving personal agent★ 246k▲ +375MITpermissivelinux · macos · windows · android · cli
DeepSeek Harness landing: blue background, the headline 'Everything is a plugin' and a quick-start box with the npx command.
DeepSeek HarnessPlugin agent harness for developers★ 225k▲ +1,356MITpermissiveweb · cli
colibrì web dashboard: hardware and VRAM/RAM/disk tiers at left, a chat with a 744B model in the centre, tok/s metrics on top.
colibriRun huge MoE models on your own PCTHIS ENTRY★ 33.8k▲ +974Apache-2.0permissivelinux · macos · windows · cli
mattpocock/skillsCARD★ 261k▲ +872MITpermissive—

ALSO WORTH SEEING

OpenHunterAILocal AI red-team workspaceOpenHunterAI's homepage, labelled "AUTHORIZED EXTERNAL SECURITY TESTING," listing its five-stage process from scope verification to reporting.Self-host
★ 165▲ +15 / 7d
An alpha-stage local workspace that runs attacker-style security assessments — strictly against public targets you own or are authorized to test.MartinLoopExecution-control CLI for coding agentsMartinLoop's homepage headline: "Watch your AI burn $5.20. Then watch MartinLoop close the loop for $2.30."
★ 130▲ +21 / 7d
Wraps a coding agent with a spend cap and a verifier, so "I'm done" comes with an evidence-backed receipt instead of just a claim.Spec KitSpec-first process for coding agentsSpec Kit docs landing page — the headline 'Build with a spec, fix a bug, or assess an idea' and the uv tool install specify-cli commandInstall to use
★ 137k▲ +560 / 7d
Makes your coding agent write the spec first, then plan, build and verify against it.HindsightNarrative map for YouTube viewersHindsight landing page: thousands of video dots grouped into issue clusters, with a central search box for a topic, creator or YouTube link.Use in the browser
★ 0▲ 1 HN
Remembers 49,411 videos, maps 22,904 by issue, traces how a YouTuber's argument changed.

TIMELINE

  1. Repository created
  2. Show HN post ↗
  3. FIRST SEEN BY TESIGN
  4. Published on TESIGN

SAME DAY

  1. 01

    VoiceStudio

    Voice cloning and dubbing on your PC

    Clones voices, dubs video and transcribes audio on your own machine.

    Self-host
    ★ 30.9k▲ +1,260 / 7d
  2. 02

    gods-eye-view

    Live satellite globe in your browser

    Puts real satellite and flight tracks on a 3D globe in your browser.

    Install to use
    ★ 34.8k▲ +2,468 / 7d
  3. 03

    Recordly

    Record and polish demo videos

    Turns a screen recording into a polished demo video without editing.

    Install to use
    ★ 29k▲ +399 / 7d
  4. 04

    Orca

    Run coding agents side by side

    Runs a fleet of coding agents in parallel from one place.

    ★ 69.5k▲ +794 / 7d
  5. 05

    archify

    System diagrams for coding-agent users

    Exports architecture and data-flow diagrams as self-contained HTML.

    For developers
    ★ 63.6k▲ +1,307 / 7d
  6. 07

    tokentab

    Cost breakdown for agent CLI users

    Reads agent session logs and works out cost by model, project and day.

    Install to use
    ★ 1.1k▲ +118 / 7d