github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #1388] Description of models in the ollama page #733

New Issue
Closed
opened 2026-04-12 10:24:22 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-12 10:24:22 -05:00
Owner
Copy Link

Originally created by @lfoppiano on GitHub (Dec 5, 2023).
Original GitHub issue: https://github.com/ollama/ollama/issues/1388

I cannot find the meaning part of the naming convention of the models. For example:

https://ollama.ai/library/starling-lm:7b-alpha-q4_K_M

It's clear that q4 indicate the bit of quantization, but what do K and M mean?

Thanks

Originally created by @lfoppiano on GitHub (Dec 5, 2023). Original GitHub issue: https://github.com/ollama/ollama/issues/1388 I cannot find the meaning part of the naming convention of the models. For example: https://ollama.ai/library/starling-lm:7b-alpha-q4_K_M It's clear that q4 indicate the bit of quantization, but what do K and M mean? Thanks
GiteaMirror commented 2026-04-12 10:24:24 -05:00
Author
Owner
Copy Link

@easp commented on GitHub (Dec 5, 2023):

S, M & L stand for small, medium and large. K refers, I think, to the quantization method, which, I think, uses K-means clustering.

<!-- gh-comment-id:1841539587 --> @easp commented on GitHub (Dec 5, 2023): S, M & L stand for small, medium and large. K refers, I think, to the quantization method, which, I think, uses K-means clustering.
GiteaMirror commented 2026-04-12 10:24:24 -05:00
Author
Owner
Copy Link

@lfoppiano commented on GitHub (Dec 5, 2023):

Thanks! but I'm wondering, for such models, what small, medium and large mean? I though that the number of parameters was the scale indication 🤔

<!-- gh-comment-id:1841807063 --> @lfoppiano commented on GitHub (Dec 5, 2023): Thanks! but I'm wondering, for such models, what small, medium and large mean? I though that the number of parameters was the scale indication 🤔
GiteaMirror commented 2026-04-12 10:24:25 -05:00
Author
Owner
Copy Link

@easp commented on GitHub (Dec 6, 2023):

Good question. I should have been more clear. The sizes are relevant to the quantization. So, within the 4-bit K-quantization there are the option of Small and Medium variants. For the 3-bit, there are S, M & L,

From: 5f6e0c0dff/examples/make-ggml.py (L16C1-L37C51)

Old quant types (some base model types require these):
- Q4_0: small, very high quality loss - legacy, prefer using Q3_K_M
- Q4_1: small, substantial quality loss - legacy, prefer using Q3_K_L
- Q5_0: medium, balanced quality - legacy, prefer using Q4_K_M
- Q5_1: medium, low quality loss - legacy, prefer using Q5_K_M

New quant types (recommended):
- Q2_K: smallest, extreme quality loss - not recommended
- Q3_K: alias for Q3_K_M
- Q3_K_S: very small, very high quality loss
- Q3_K_M: very small, very high quality loss
- Q3_K_L: small, substantial quality loss
- Q4_K: alias for Q4_K_M
- Q4_K_S: small, significant quality loss
- Q4_K_M: medium, balanced quality - recommended
- Q5_K: alias for Q5_K_M
- Q5_K_S: large, low quality loss - recommended
- Q5_K_M: large, very low quality loss - recommended
- Q6_K: very large, extremely low quality loss
- Q8_0: very large, extremely low quality loss - not recommended
- F16: extremely large, virtually no quality loss - not recommended
- F32: absolutely huge, lossless - not recommended
<!-- gh-comment-id:1841890505 --> @easp commented on GitHub (Dec 6, 2023): Good question. I should have been more clear. The sizes are relevant to the quantization. So, within the 4-bit K-quantization there are the option of Small and Medium variants. For the 3-bit, there are S, M & L, From: https://github.com/ggerganov/llama.cpp/blob/5f6e0c0dff1e7a89331e6b25eca9a9fd71324069/examples/make-ggml.py#L16C1-L37C51 ``` Old quant types (some base model types require these): - Q4_0: small, very high quality loss - legacy, prefer using Q3_K_M - Q4_1: small, substantial quality loss - legacy, prefer using Q3_K_L - Q5_0: medium, balanced quality - legacy, prefer using Q4_K_M - Q5_1: medium, low quality loss - legacy, prefer using Q5_K_M New quant types (recommended): - Q2_K: smallest, extreme quality loss - not recommended - Q3_K: alias for Q3_K_M - Q3_K_S: very small, very high quality loss - Q3_K_M: very small, very high quality loss - Q3_K_L: small, substantial quality loss - Q4_K: alias for Q4_K_M - Q4_K_S: small, significant quality loss - Q4_K_M: medium, balanced quality - recommended - Q5_K: alias for Q5_K_M - Q5_K_S: large, low quality loss - recommended - Q5_K_M: large, very low quality loss - recommended - Q6_K: very large, extremely low quality loss - Q8_0: very large, extremely low quality loss - not recommended - F16: extremely large, virtually no quality loss - not recommended - F32: absolutely huge, lossless - not recommended ```
GiteaMirror commented 2026-04-12 10:24:25 -05:00
Author
Owner
Copy Link

@lfoppiano commented on GitHub (Dec 6, 2023):

I understand, thanks.

<!-- gh-comment-id:1841932262 --> @lfoppiano commented on GitHub (Dec 6, 2023): I understand, thanks.
GiteaMirror commented 2026-04-12 10:24:26 -05:00
Author
Owner
Copy Link

@b-a0 commented on GitHub (Jul 25, 2024):

And is there a difference between upper and lower case letters?

image

<!-- gh-comment-id:2250831755 --> @b-a0 commented on GitHub (Jul 25, 2024): And is there a difference between upper and lower case letters? ![image](https://github.com/user-attachments/assets/b8a450e5-8cfe-447e-8a56-a9b68eeaa74d)
GiteaMirror commented 2026-04-12 10:24:28 -05:00
Author
Owner
Copy Link

@github12101 commented on GitHub (Jan 27, 2025):

Good question!

<!-- gh-comment-id:2617092250 --> @github12101 commented on GitHub (Jan 27, 2025): Good question!
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#733
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?