Originally created by @qhjqhj on GitHub (Mar 6, 2025).
Check Existing Issues
I have searched the existing issues and discussions.
Installation Method
Git Clone
Open WebUI Version
0.5.20
Ollama Version (if applicable)
0.5.13
Operating System
windows 11
Browser (if applicable)
No response
Confirmation
I have read and followed all instructions in README.md.
I am using the latest version of both Open WebUI and Ollama.
I have checked the browser console logs.
I have checked the Docker container logs.
I have listed steps to reproduce the bug in detail.
Expected Behavior
Separate thinking from formal output
Actual Behavior
The content of thinking and the content of formal output cannot be distinguished, and the formal content cannot be output
Steps to Reproduce
run ollama serve
run open-webui serve
Open the browser and select the qwq model
Ask a code-related question, such as "Write the code for the Viterbi algorithm using Python"
When running the qwq:latest model of Ollama on the backend, the content in the tag is output as normal content, and part of the code output seems to be lost
Logs & Screenshots
Additional Information
No response
Originally created by @qhjqhj on GitHub (Mar 6, 2025).
### Check Existing Issues
- [x] I have searched the existing issues and discussions.
### Installation Method
Git Clone
### Open WebUI Version
0.5.20
### Ollama Version (if applicable)
0.5.13
### Operating System
windows 11
### Browser (if applicable)
_No response_
### Confirmation
- [x] I have read and followed all instructions in `README.md`.
- [x] I am using the latest version of **both** Open WebUI and Ollama.
- [x] I have checked the browser console logs.
- [x] I have checked the Docker container logs.
- [x] I have listed steps to reproduce the bug in detail.
### Expected Behavior
Separate thinking from formal output
### Actual Behavior
The content of thinking and the content of formal output cannot be distinguished, and the formal content cannot be output
### Steps to Reproduce
1. run ollama serve
2. run open-webui serve
3. Open the browser and select the qwq model
4. Ask a code-related question, such as "Write the code for the Viterbi algorithm using Python"
When running the qwq:latest model of Ollama on the backend, the content in the <think> tag is output as normal content, and part of the code output seems to be lost
### Logs & Screenshots


### Additional Information
_No response_
GiteaMirror
added the bug label 2025-11-11 15:50:18 -06:00
Adding to this, it doesn't even seem to have its normal output in the thinking tag.
On my end, everything in the thinking tag seems to indeed be the thought process, but once it's done it outputs no answer, and openwebui acts as if the model was still thinking.
@Eidenz commented on GitHub (Mar 6, 2025):
Adding to this, it doesn't even seem to have its normal output in the thinking tag.
On my end, everything in the thinking tag seems to indeed be the thought process, but once it's done it outputs no answer, and openwebui acts as if the model was still thinking.
It seems that the model's thinking output has been stopped before it has finished, please try increasing nun_ctx and num_predict
@EntropyYue commented on GitHub (Mar 6, 2025):
It seems that the model's thinking output has been stopped before it has finished, please try increasing `nun_ctx` and `num_predict`
It seems that the model's thinking output has been stopped before it has finished, please try increasing nun_ctx and num_predict
I tried it several times, and the total number of tokens generated was different. It seems that the issue is not caused by the maximum length limit.
@qhjqhj commented on GitHub (Mar 6, 2025):
> It seems that the model's thinking output has been stopped before it has finished, please try increasing `nun_ctx` and `num_predict`
I tried it several times, and the total number of tokens generated was different. It seems that the issue is not caused by the maximum length limit.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @qhjqhj on GitHub (Mar 6, 2025).
Check Existing Issues
Installation Method
Git Clone
Open WebUI Version
0.5.20
Ollama Version (if applicable)
0.5.13
Operating System
windows 11
Browser (if applicable)
No response
Confirmation
README.md.Expected Behavior
Separate thinking from formal output
Actual Behavior
The content of thinking and the content of formal output cannot be distinguished, and the formal content cannot be output
Steps to Reproduce
When running the qwq:latest model of Ollama on the backend, the content in the tag is output as normal content, and part of the code output seems to be lost
Logs & Screenshots
Additional Information
No response
@Eidenz commented on GitHub (Mar 6, 2025):
Adding to this, it doesn't even seem to have its normal output in the thinking tag.
On my end, everything in the thinking tag seems to indeed be the thought process, but once it's done it outputs no answer, and openwebui acts as if the model was still thinking.
@mihkhub commented on GitHub (Mar 6, 2025):
QwQ-32b thinking tag pair "
@EntropyYue commented on GitHub (Mar 6, 2025):
It seems that the model's thinking output has been stopped before it has finished, please try increasing
nun_ctxandnum_predict@Duxon commented on GitHub (Mar 6, 2025):
How to increase
num_ctx(to 131072) in the Open-WebUI settings? I can't find that seeting.@EntropyYue commented on GitHub (Mar 6, 2025):
If you are using en-US, it will display as
Context Length@qhjqhj commented on GitHub (Mar 6, 2025):
I tried it several times, and the total number of tokens generated was different. It seems that the issue is not caused by the maximum length limit.
@bjj commented on GitHub (Mar 6, 2025):
See #11259
@EntropyYue commented on GitHub (Mar 6, 2025):
Please provide screenshots of other situations, which will help us better solve the problem