github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-07-16 13:19:40 -05:00
Code Issues 16.9k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #11405] issue with disable thinking in quantized variants #7527

New Issue
Closed
opened 2026-04-12 19:37:41 -05:00 by GiteaMirror · 9 comments
GiteaMirror commented 2026-04-12 19:37:41 -05:00
Owner
Copy Link

Originally created by @Abdulrahman392011 on GitHub (Jul 13, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11405

What is the issue?

i noticed that most of the other quantization variants of the same thinking model doesn't have the disable thinking feature.

can you add it so that any model that contain the main name of the model that is registered as thinking model is treated as thinking model. currently it has to be the exact name that is registered as thinking, that's why it doesn't treat the quantized variants as thinking models.

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

No response

Originally created by @Abdulrahman392011 on GitHub (Jul 13, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11405 ### What is the issue? i noticed that most of the other quantization variants of the same thinking model doesn't have the disable thinking feature. can you add it so that any model that contain the main name of the model that is registered as thinking model is treated as thinking model. currently it has to be the exact name that is registered as thinking, that's why it doesn't treat the quantized variants as thinking models. ### Relevant log output ```shell ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version _No response_
GiteaMirror added the bug label 2026-04-12 19:37:41 -05:00
GiteaMirror commented 2026-04-12 19:37:43 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jul 13, 2025):

Which model?

<!-- gh-comment-id:3067159566 --> @rick-github commented on GitHub (Jul 13, 2025): Which model?
GiteaMirror commented 2026-04-12 19:37:44 -05:00
Author
Owner
Copy Link

@Davidyz commented on GitHub (Jul 14, 2025):

I ran into a similar problem when I tried to use a qwen3 gguf from unsloth instead of ollama registry. Fixed it by creating a modelfile that load the unsloth gguf with the ollama chat template. I think this should be better documented (ie. how to make a reasoning model work with the think parameter).

<!-- gh-comment-id:3068044162 --> @Davidyz commented on GitHub (Jul 14, 2025): I ran into a similar problem when I tried to use a qwen3 gguf from unsloth instead of ollama registry. Fixed it by creating a modelfile that load the unsloth gguf with the ollama chat template. I think this should be better documented (ie. how to make a reasoning model work with the `think` parameter).
GiteaMirror commented 2026-04-12 19:37:44 -05:00
Author
Owner
Copy Link

@Abdulrahman392011 commented on GitHub (Jul 18, 2025):

@rick-github

use any model that has a quantization variant in it's name.

like @Davidyz said ollama uses it's model file with chat template to recognize which model is thinking and which is not. which is leaving every other model and variant of the same model unrecognized cause their model file weren't created with ollama chat template.

using the model file is too complicated of a strategy. it will be much easier to make a list of names that if recognized to be in the model name, it will deal with it as a thinking model.

and just add the names of the thinking models to that list.

<!-- gh-comment-id:3090300076 --> @Abdulrahman392011 commented on GitHub (Jul 18, 2025): @rick-github use any model that has a quantization variant in it's name. like @Davidyz said ollama uses it's model file with chat template to recognize which model is thinking and which is not. which is leaving every other model and variant of the same model unrecognized cause their model file weren't created with ollama chat template. using the model file is too complicated of a strategy. it will be much easier to make a list of names that if recognized to be in the model name, it will deal with it as a thinking model. and just add the names of the thinking models to that list.
GiteaMirror commented 2026-04-12 19:37:45 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jul 18, 2025):

Since you don't want to supply the model name, I assume you are importing a model. When importing a model, a template needs to be provided. Ollama can detect and automatically add a template for some models, but if that fails, it's up to the user to provide an appropriate template. The user can use a template from a similar model in the Ollama library as a starting point.

<!-- gh-comment-id:3090765327 --> @rick-github commented on GitHub (Jul 18, 2025): Since you don't want to supply the model name, I assume you are importing a model. When importing a model, a template needs to be provided. Ollama can detect and automatically add a template for some models, but if that fails, it's up to the user to provide an appropriate template. The user can use a template from a similar model in the Ollama library as a starting point.
GiteaMirror commented 2026-04-12 19:37:45 -05:00
Author
Owner
Copy Link

@Abdulrahman392011 commented on GitHub (Jul 18, 2025):

i tried ollama run qwen3:30b-a3b it won't work but i tried other models as well, same issue.

i understand what you're saying. it just feels like the normal user won't be able to do that. but your guide lines are perfect for the problem.

so we can close this issue it's not that big of a problem anyways. why would you want to disable thinking if it's a thinking model, right!

<!-- gh-comment-id:3090865250 --> @Abdulrahman392011 commented on GitHub (Jul 18, 2025): i tried ollama run qwen3:30b-a3b it won't work but i tried other models as well, same issue. i understand what you're saying. it just feels like the normal user won't be able to do that. but your guide lines are perfect for the problem. so we can close this issue it's not that big of a problem anyways. why would you want to disable thinking if it's a thinking model, right!
GiteaMirror commented 2026-04-12 19:37:46 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jul 18, 2025):

qwen3:30b-a3b supports thinking:

$ ollama run qwen3:30b-a3b-q4_K_M hello
Thinking...
Okay, the user just said "hello". I need to respond appropriately. Let me think about the best way to greet them back.

First, I should keep it friendly and welcoming. Maybe start with a cheerful greeting like "Hello!" to match their tone.

Next, I should ask how I can assist them. That way, they know I'm here to help with whatever they need. Something like, "How can I assist you today?" would be good.

I should also make sure the response is concise and not too wordy. No need for extra fluff, just a simple and polite reply.

Wait, maybe add an emoji to keep it friendly? Like a smiley face. But I should check if that's appropriate. The user didn't use any, but it might make the response feel more personable.

Putting it all together: "Hello! 😊 How can I assist you today?" That sounds good. It's friendly, open, and invites them to ask for help. I think that's a solid response.
...done thinking.

Hello! 😊 How can I assist you today?

Have you tried re-pulling the model to make sure it's up to date?

why would you want to disable thinking if it's a thinking model, right!

Some users prefer a quicker response, skipping the thinking allows that.

<!-- gh-comment-id:3091110255 --> @rick-github commented on GitHub (Jul 18, 2025): qwen3:30b-a3b supports thinking: ```console $ ollama run qwen3:30b-a3b-q4_K_M hello Thinking... Okay, the user just said "hello". I need to respond appropriately. Let me think about the best way to greet them back. First, I should keep it friendly and welcoming. Maybe start with a cheerful greeting like "Hello!" to match their tone. Next, I should ask how I can assist them. That way, they know I'm here to help with whatever they need. Something like, "How can I assist you today?" would be good. I should also make sure the response is concise and not too wordy. No need for extra fluff, just a simple and polite reply. Wait, maybe add an emoji to keep it friendly? Like a smiley face. But I should check if that's appropriate. The user didn't use any, but it might make the response feel more personable. Putting it all together: "Hello! 😊 How can I assist you today?" That sounds good. It's friendly, open, and invites them to ask for help. I think that's a solid response. ...done thinking. Hello! 😊 How can I assist you today? ``` Have you tried re-pulling the model to make sure it's up to date? > why would you want to disable thinking if it's a thinking model, right! Some users prefer a quicker response, skipping the thinking allows that.
GiteaMirror commented 2026-04-12 19:37:46 -05:00
Author
Owner
Copy Link

@Abdulrahman392011 commented on GitHub (Jul 18, 2025):

@rick-github you're right the model has been sitting in my laptop since it was out. I will try repulling it. That would surly fix the issue for me.

But have you tried disabling thinking that is usually the part that fails. Even if the model doesn't get recognised as thinking model it still works as thinking normally

<!-- gh-comment-id:3091211572 --> @Abdulrahman392011 commented on GitHub (Jul 18, 2025): @rick-github you're right the model has been sitting in my laptop since it was out. I will try repulling it. That would surly fix the issue for me. But have you tried disabling thinking that is usually the part that fails. Even if the model doesn't get recognised as thinking model it still works as thinking normally
GiteaMirror commented 2026-04-12 19:37:47 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jul 18, 2025):

$ ollama run qwen3:30b-a3b-q4_K_M hello --think=false
Hello! How can I assist you today?
<!-- gh-comment-id:3091225954 --> @rick-github commented on GitHub (Jul 18, 2025): ```console $ ollama run qwen3:30b-a3b-q4_K_M hello --think=false Hello! How can I assist you today? ```
GiteaMirror commented 2026-04-12 19:37:48 -05:00
Author
Owner
Copy Link

@Abdulrahman392011 commented on GitHub (Jul 19, 2025):

I repulled it and it worked. thanks a lot @rick-github

<!-- gh-comment-id:3092562287 --> @Abdulrahman392011 commented on GitHub (Jul 19, 2025): I repulled it and it worked. thanks a lot @rick-github
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 bug
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#7527
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?