Models

Kolibri 1

Aleph-Alpha · October 2026

Get

Hugging Face

A 78B, 3.8B active chat model from Aleph-Alpha.

Downloads: 65K
Likes: 35
Parameters: 78B, 3.8B active
Context: 256K
Licence: apache-2.0

Too big for this Mac

Its smallest size needs 21 GB of memory, more than M4 · 16 GB can give it.

Kolibri 1 runs well from a Mac with an M1 Pro and 32 GB of memory. 40 of 59 Mac configurations run it well.

Compare with

The nearest models in size.

Sizes

From bartowski/Kolibri-1-GGUF on Hugging Face. A smaller size is a smaller download and a little less accurate.

Size	Download	Memory at 8K	Get
IQ1_S	18 GB	18 GBestimated	Open
IQ1_M	20 GB	20 GBestimated	Open
IQ2_XXS	23 GB	23 GBestimated	Open
IQ2_XS	24 GB	25 GBestimated	Open
IQ2_S	26 GB	26 GBestimated	Open
IQ2_M	28 GB	28 GBestimated	Open
Q2_K	30 GB	30 GBestimated	Open
IQ3_XXS	34 GB	34 GBestimated	Open
IQ3_XS	35 GB	36 GBestimated	Open
Q3_K_S	36 GB	36 GBestimated	Open
Q3_K_M	38 GB	38 GBestimated	Open
Q3_K_L	40 GB	40 GBestimated	Open
IQ3_M	41 GB	41 GBestimated	Open
IQ4_XS	44 GB	44 GBestimated	Open
Q4_0	45 GB	45 GBestimated	Open
Q4_K_S	46 GB	47 GBestimated	Open
Q4_1	49 GB	49 GBestimated	Open
IQ4_NL	50 GB	50 GBestimated	Open
Q4_K_M	50 GB	50 GBestimated	Open
Q4_K_L	54 GB	54 GBestimated	Open
Q5_K_S	56 GB	56 GBestimated	Open
Q5_K_M	60 GB	61 GBestimated	Open
Q6_K_S	66 GB	66 GBestimated	Open
Q6_K	69 GB	69 GBestimated	Open
Q6_K_L	73 GB	73 GBestimated	Open
Q8_0	83 GB	83 GBestimated	Open
BF16	156 GB	157 GBestimated	Open

Run It Anywhere

Kolibri 1 is an open model, so any app that runs GGUF files can run it. These take the IQ1_S size from bartowski straight from Hugging Face.

LM StudioMac, Windows, LinuxOpen
JanMac, Windows, LinuxOpen
MstyMac, Windows, LinuxOpen
OllamaMac, Windows, Linuxollama run hf.co/bartowski/Kolibri-1-GGUF:IQ1_S
llama.cppMac, Windows, Linuxllama-server -hf bartowski/Kolibri-1-GGUF:IQ1_S
Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/bartowski/Kolibri-1-GGUF:IQ1_S

On a PC Graphics Card

Video memory	Largest size that fits	Download	Memory at 8K
6 GB	None worth running		needs 34 GB
8 GB	None worth running		needs 34 GB
10 GB	None worth running		needs 34 GB
11 GB	None worth running		needs 34 GB
12 GB	None worth running		needs 34 GB
16 GB	None worth running		needs 34 GB
20 GB	None worth running		needs 34 GB
24 GB	None worth running		needs 34 GB
32 GB	None worth running		needs 34 GB
48 GB	Q4_K_S	46 GB	47 GB

The largest size, of those that keep three bits a weight or more, whose memory at an 8K chat is under 92% of the card's; the rest is the desktop's. Estimated from the files, the same way as for a Mac; speed on a PC is not estimated yet. A size that doesn't fit the card still runs, slower, with part of it in ordinary memory.

On every Mac

Mac	Size	Fit	Memory	Writes
M6 · 16 GB	IQ1_S	Too big	21 GB
M6 · 24 GB	IQ1_S	Tight fit	19 GB	36
M6 · 32 GB	IQ1_M	Runs well	21 GB	34
M5 · 16 GB	IQ1_S	Too big	21 GB
M5 · 24 GB	IQ1_S	Tight fit	19 GB	35
M5 · 32 GB	IQ1_M	Runs well	21 GB	33
M5 Pro · 24 GB	IQ1_S	Tight fit	19 GB	61
M5 Pro · 48 GB	Q2_K	Runs well	32 GB	47
M5 Pro · 64 GB	IQ3_M	Runs well	42 GB	39
M5 Max · 36 GB	IQ2_XXS	Runs well	24 GB	72
M5 Max · 48 GB	Q2_K	Runs well	32 GB	78
M5 Max · 64 GB	IQ3_M	Runs well	42 GB	67
M5 Max · 128 GB	Q6_K_S	Runs great	68 GB	52
M5 Ultra · 96 GB	Q4_0	Runs great	46 GB	113
M5 Ultra · 256 GB	Q8_0	Runs great	86 GB	85
M5 Ultra · 512 GB	Q8_0	Runs great	86 GB	85
M4 · 16 GB	IQ1_S	Too big	21 GB
M4 · 24 GB	IQ1_S	Tight fit	19 GB	21
M4 · 32 GB	IQ1_M	Runs well	21 GB	20
M4 Pro · 24 GB	IQ1_S	Tight fit	19 GB	33
M4 Pro · 48 GB	Q2_K	Runs well	32 GB	28
M4 Pro · 64 GB	IQ3_M	Runs well	42 GB	25
M4 Max · 36 GB	IQ2_XXS	Runs well	24 GB	37
M4 Max · 48 GB	Q2_K	Runs well	32 GB	38
M4 Max · 64 GB	IQ3_M	Runs well	42 GB	35
M4 Max · 96 GB	Q4_0	Runs great	46 GB	34
M4 Max · 128 GB	Q6_K_S	Runs great	68 GB	30
M3 · 8 GB	IQ1_S	Too big	21 GB
M3 · 16 GB	IQ1_S	Too big	21 GB
M3 · 24 GB	IQ1_S	Tight fit	19 GB	20
M3 Pro · 18 GB	IQ1_S	Too big	21 GB
M3 Pro · 36 GB	IQ2_XXS	Runs well	24 GB	23
M3 Max · 36 GB	IQ2_XXS	Runs well	24 GB	35
M3 Max · 48 GB	Q2_K	Runs well	32 GB	37
M3 Max · 64 GB	IQ3_M	Runs well	42 GB	33
M3 Max · 96 GB	Q4_0	Runs great	46 GB	31
M3 Max · 128 GB	Q6_K_S	Runs great	68 GB	27
M3 Ultra · 96 GB	Q4_0	Runs great	46 GB	50
M3 Ultra · 256 GB	Q8_0	Runs great	86 GB	40
M3 Ultra · 512 GB	Q8_0	Runs great	86 GB	40
M2 · 8 GB	IQ1_S	Too big	21 GB
M2 · 16 GB	IQ1_S	Too big	21 GB
M2 · 24 GB	IQ1_S	Tight fit	19 GB	20
M2 Pro · 16 GB	IQ1_S	Too big	21 GB
M2 Pro · 32 GB	IQ1_M	Runs well	21 GB	29
M2 Max · 32 GB	IQ1_M	Runs well	21 GB	41
M2 Max · 64 GB	IQ3_M	Runs well	42 GB	33
M2 Max · 96 GB	Q4_0	Runs great	46 GB	36
M2 Ultra · 64 GB	IQ3_M	Runs well	42 GB	44
M2 Ultra · 128 GB	Q6_K_S	Runs great	68 GB	38
M2 Ultra · 192 GB	Q8_0	Runs great	86 GB	40
M1 · 8 GB	IQ1_S	Too big	21 GB
M1 · 16 GB	IQ1_S	Too big	21 GB
M1 Pro · 16 GB	IQ1_S	Too big	21 GB
M1 Pro · 32 GB	IQ1_M	Runs well	21 GB	23
M1 Max · 32 GB	IQ1_M	Runs well	21 GB	30
M1 Max · 64 GB	IQ3_M	Runs well	42 GB	25
M1 Ultra · 64 GB	IQ3_M	Runs well	42 GB	31
M1 Ultra · 128 GB	Q6_K_S	Runs great	68 GB	28

Writes is words a second, estimated from the Mac's memory speed and the model's size; how, and how wrong.

Packaged by

5 repositories on Hugging Face carry this model for llama.cpp. Outmake reads them all as one model.

Information

Maker: Aleph-Alpha
Released: October 2026
Parameters: 78B, 3.8B active
Longest chat: 256K tokens
Reads pictures: No
Licence: apache-2.0
Downloads this month: 65,108
Likes: 35
Sizes: 27, 18 to 156 GB
Kind: Maker's release
Source: Aleph-Alpha/Kolibri-1

Badge for the model card

Shows on Kolibri 1's Hugging Face card which Macs run it well, linked to this page. It updates every morning.

Files, downloads and likes read from Hugging Face on November 5, 2026 and refreshed every morning. Memory and speed are estimates until someone running Outmake measures this model on that Mac.