colibri
AI
local llm inference
SHOW HN2026
How HN took it
32 positive, 81 neutral or asking, 4 mixed, 10 critical, out of 127 comments about the project. Thread score +0.18.
Performance: 6 for, 8 againstIdea: 12 for, 0 againstExecution: 2 for, 0 againstPrivacy: 1 for, 0 againstPricing: 0 for, 1 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Impressive technical achievement running a huge model on modest hardware
From 3 comments
This is great, well done! I love seeing people run things where they weren't meant to be run.
stavros on HNWell under 1tps, but the fact it runs at all on a 16gb system is incredibly impressive!
AussieWog93 on HNYeah this idea makes instant sense. Very well done, this deserves a github star on concept alone.
barent on HN
Fits the hacker spirit and an interesting idea
From 2 comments
This is the hacker spirit
miohtama on HNThe best ideas are the ones that seem obvious. This is one of those ideas.
barent on HN
Objections
Speed around 0.05 to 1 token per second is too slow for practical use
From 3 comments
My main question is whether when put into practical use, this can be measured in tokens/second, or more like 1 token per minute... I have seen locally hosted LLM that are as slow…
walrus01 on HNThis is technically impressive, but is it usable in practice?
3abiton on HN"Slow computer" + running a 200B model = "I'll just wait a week."
zftnb666 on HN
Frequent SSD reads may accelerate wear on drives
From 3 comments
The page has an SSD wear warning [0] I use desktop PCs that I build from components so I can replace the SSD, but what do users with soldered SSD do? Just avoid these…
voidmain0001 on HNWould this cause issues with SSD lifespan?
Pragmata on HNis this bad for my SSD?
doodlesarefun on HN
Requests
Support for multiple disks or RAID to speed up reads
From 3 comments
I am curious if it's possible to adjust this to use more RAM, as i've got a machine with 64GB RAM and 24GB VRAM. Or perhaps I could run Gemma/Qwen on the GPU and have GLM-5.2…
tarpitt on HNI wonder how would a RAID0 array of either disks or even nvme improve the performance of this.
xtracto on HNThis is great stuff, I feel like fast disk (SSD) is a somewhat solvable problem, if you have many disks with the same content and a fast controller (?).
webprofusion on HN
Energy efficiency benchmarks
From 1 comment
I’m truly impressed by your work ! I don’t know if this is planned for near future, but how about adding energy efficiency benchmarks ? Because running locally is a great…
maxignol on HN
Questions
Can it use more RAM or a GPU alongside SSD streaming
From 1 comment
I am curious if it's possible to adjust this to use more RAM, as i've got a machine with 64GB RAM and 24GB VRAM. Or perhaps I could run Gemma/Qwen on the GPU and have GLM-5.2…
tarpitt on HN
Could it run on Ubuntu as a coding agent
From 1 comment
I am curious if it's possible to use coli on ubuntu as coding agent, as i've got a machine with 32GB RAM and 2T SSD.
kelvinxue on HN
Alternatives named
llama.cpp with mmap
From 1 comment
llama.cpp supports a wide variety of 4-bit and smaller quants and mmap's models by default, so you dont need to be able to hold the weights in memory (the OS will handle bringing…
ac29 on HN
flash-moe
From 1 comment
Another recent project that runs a huge model on a 48gb Mac is https://github.com/danveloper/flash-moe - it gets over 5 tokens/sec on an M3 Max compared to this projects very…
bahmboo on HN
fastllm
From 1 comment
Is this similar to fastllm? https://github.com/ztxz16/fastllm
nogajun on HN
ds4
From 1 comment
Looking at the progress with DS4 project from Antirez and Colibri, are we probably going in the direction of having a stripe RAID of SSDs instead of RAM to get decent performance…
rullopat on HN