Colibri
Ranked 11 of 27 local LLM runners by recommenders
Recommended for running large models from disk on modest hardware and as a Discord-like chat with nice UX; critics say its privacy falls short of its marketing and that local inference is very slow.
Recommended for running large models from disk on modest hardware and as a Discord-like chat with nice UX; critics say its privacy falls short of its marketing and that local inference is very slow.
33 people gave an opinion
- recommend it
- 20
- criticize it
- 13
- approval share
- 61%
Recommended in 3 of 13 months
- November 2025: not recommended
- December 2025: not recommended
- January 2026: not recommended
- February 2026: not recommended
- March 2026: recommended
- April 2026: not recommended
- May 2026: not recommended
- June 2026: not recommended
- July 2026: recommended
- August 2026: recommended
- September 2026: not recommended
- October 2026: not recommended
- November 2026: not recommended (month in progress)
Why it doesn’t carry the badge
- 20 people recommend itNeeds 10 or more, met.
- 1.5 recommenders for every criticNeeds 2 to 1 or better, not met.
- Recommended in 3 of 13 monthsNeeds 4 or more, not met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Streams model weights from disk
“Already is possible. On a machine with 32GB ram, and NO gpu. Just need a large SSD or NVME. Streams from disk to memory.”
Comment from 21 Jul 2026 in “Who's Afraid of Chinese Models?” “There are more focused frameworks like DS4 and Colibri which have better support for streaming weights from disk”
Comment from 14 Aug 2026 in “GLM-5.3: Frontier Coding with Emergent Cyber Capabilities” Discord alternative with nice UX
“UX is nice and should encourage people to switch from Discord”
Comment from 26 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small” “Nice to see more Discord alternatives these days.”
Comment from 26 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small” Promising and cool project
“This project seems more of an experiment than something everyone should run, but pretty cool nonetheless”
Comment from 9 Jul 2026 in “Getting GLM 5.2 running on my slow computer” “This looks very promising!”
Comment from 29 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small”
What people warn about
Each mark is one comment
Messages not private or end-to-end encrypted
“It's a nonstarter without E2EE for my use-cases, unfortunately.”
Comment from 27 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small” “By using ATProto, Colibri fundamentally makes all of your communication within any community completely public to everyone on the internet.”
Comment from 26 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small” “Colibri's marketing around ATProto creates an implied trust that doesn't actually exist at the privacy level.”
Comment from 27 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small” - Comment from 26 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small”
- Comment from 26 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small”
- Comment from 26 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small”
Very slow token generation
“It's a great concept but I think it would cross the line from 'very slow' to 'so slow it's unusable' at this size.”
Comment from 27 Jul 2026 in “Kimi-K3 Releases on HuggingFace 6/27” “Note the performance stories - 0.07 token/s on a 24 GB laptop with 24 CPU threads!”
Comment from 11 Jul 2026 in “Getting GLM 5.2 running on my slow computer” “they are doing similar things with bigger models, but tok/s is not high”
Comment from 29 Jul 2026 in “Show HN: Open-source engine running Gemma 4 26B in 2 GB RAM on any M-series Mac” - Comment from 27 Jul 2026 in “Kimi-K3 Releases on HuggingFace 6/27”
Privacy marketing overstates protection
“I feel like Colibri (based on their page) just doesn't believe it's there problem, which seems... irresponsible.”
Comment from 27 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small” “Colibri's marketing around ATProto creates an implied trust that doesn't actually exist at the privacy level.”
Comment from 27 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small” “It's disingenuous to even mention this in the marketing copy.”
Comment from 26 Mar 2026 in “Colibri – chat platform built on the AT Protocol for communities big and small”
Other local LLM runners people recommend
The four most recommended other local LLM runners, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
| Product | Recommend | Criticize | Approval | Months | Owned | Badge |
|---|---|---|---|---|---|---|
| llama.cpp | 471 | 109 | 81% | 13 | 1 yr4 owners | Approved |
| Ollama | 350 | 247 | 59% | 12 | 1.5 yrs8 owners | — |
| LM Studio | 217 | 74 | 75% | 12 | 3 yrs4 owners | Approved |
| DS4 by antirez | 109 | 20 | 84% | 7 | 1 mo3 owners | Approved |
| ColibriThis page | 20 | 13 | 61% | 3 | — | — |
What people said
34 of the 36 comments that mention it, newest first. Each links to the original comment.
“even if you're smart enough to know that colibri is doing something that's particularly applicable”
“There are more focused frameworks like DS4 and Colibri which have better support for streaming weights from disk”
“they are doing similar things with bigger models, but tok/s is not high”
“I feel like it all started after colibri [1] was released.”
“It's a great concept but I think it would cross the line from 'very slow' to 'so slow it's unusable' at this size.”
“so far I've only seen extremely slow performance”
“Painfully slow tok/s though.”
“Already is possible. On a machine with 32GB ram, and NO gpu. Just need a large SSD or NVME. Streams from disk to memory.”
“It says so right in the readme.”
“Note the performance stories - 0.07 token/s on a 24 GB laptop with 24 CPU threads!”
“The application does not perform writes, so the application doesn't actually wear your SSD at all.”
“This project seems more of an experiment than something everyone should run, but pretty cool nonetheless”
“forge ahead disregarding the possible early burnout of their storage”
“I have to say I've seen great times!”
“This looks very promising!”
“Colibri has poor messaging and understanding”
“I feel like Colibri (based on their page) just doesn't believe it's there problem, which seems... irresponsible.”
“Not sure who the target audience for this is?”
“It's a nonstarter without E2EE for my use-cases, unfortunately.”
“"For public private conversations" Go home guys”
“Colibri's marketing around ATProto creates an implied trust that doesn't actually exist at the privacy level.”
“No need to have access to posts if you aren't touching them.”
“colibri does not present itself as a discord alternative”
“Colibri is implemented in the least privacy preserving way possible”
“every app needs a new account which kind of defeats many of the interop promises, if apps even allow it (colibri requires invite code)”
“By using ATProto, Colibri fundamentally makes all of your communication within any community completely public to everyone on the internet.”
“It's a very cool product but you have to let people know their messages aren't private.”
“Nice to see more Discord alternatives these days.”
“feel free to DM me (@colibri.social) on Bluesky if you want to migrate to the Colibri PDS!”
“We do host one ourselves.”
“UX is nice and should encourage people to switch from Discord”
“to only require as few permissions as needed”
““Your data isn’t trapped on our servers” - where is it then? Who can access it?”
“It's disingenuous to even mention this in the marketing copy.”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.