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 #13193] More DeepSeek-OCR quantizations #8722

New Issue
Closed
opened 2026-04-12 21:29:35 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-04-12 21:29:35 -05:00
Owner
Copy Link

Originally created by @th1nhhdk on GitHub (Nov 21, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/13193

Hello, I would like for DeepSeek-OCR on Ollama Website to have more quantizations, such as Q4_K_M, Q6_K_M, ... currently F16 is too big for some low end systems

If you don't want to quantizatize then can you at least tell me how to do it myself? Thanks

Originally created by @th1nhhdk on GitHub (Nov 21, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/13193 Hello, I would like for DeepSeek-OCR on Ollama Website to have more quantizations, such as Q4_K_M, Q6_K_M, ... currently F16 is too big for some low end systems If you don't want to quantizatize then can you at least tell me how to do it myself? Thanks
GiteaMirror commented 2026-04-12 21:29:36 -05:00
Author
Owner
Copy Link

@maternion commented on GitHub (Nov 21, 2025):

They had higher quantizations, but according to pdevine they started giving bad results and were of no good. Maybe they will fix it and re-push those quants.

<!-- gh-comment-id:3562889895 --> @maternion commented on GitHub (Nov 21, 2025): They had higher quantizations, but according to pdevine they started giving bad results and were of no good. Maybe they will fix it and re-push those quants.
GiteaMirror commented 2026-04-12 21:29:37 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 21, 2025):

The more quantized models worked, it's just the that results weren't quite as good due to the loss of precision. You can quantize to q4 if you want to try:

echo FROM deepseek-ocr:3b-bf16 > Modelfile
ollama create deepseek-ocr:3b-q4_K_M -q q4_K_M
<!-- gh-comment-id:3563006985 --> @rick-github commented on GitHub (Nov 21, 2025): The more quantized models worked, it's just the that results weren't quite as good due to the loss of precision. You can quantize to q4 if you want to try: ```console echo FROM deepseek-ocr:3b-bf16 > Modelfile ollama create deepseek-ocr:3b-q4_K_M -q q4_K_M ```
GiteaMirror commented 2026-04-12 21:29:37 -05:00
Author
Owner
Copy Link

@th1nhhdk commented on GitHub (Nov 21, 2025):

Yes, I have tried both Q4_K_M and Q8_0:

  • Q4_K_M gives disappointing results
  • Q8_0 stuck in a infinite loop, doesn't work at all
<!-- gh-comment-id:3563086135 --> @th1nhhdk commented on GitHub (Nov 21, 2025): Yes, I have tried both Q4_K_M and Q8_0: - Q4_K_M gives disappointing results - Q8_0 stuck in a infinite loop, doesn't work at all
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#8722
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?