github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 08:30:05 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #15363] The qwen3.5 35b model has a bug that causes infinite thinking when using the network. #71890

New Issue
Open
opened 2026-05-05 02:52:09 -05:00 by GiteaMirror · 4 comments
GiteaMirror commented 2026-05-05 02:52:09 -05:00
Owner
Copy Link

Originally created by @sdqq1234 on GitHub (Apr 6, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15363

What is the issue?

I'm using Ollama version 0.20.2 to load a Q4 quantization model of QW3.5 35B. When I connect to the internet and provide a image to query its content, I encounter an infinite "thinking" loop where the content keeps looping without ending.

Relevant log output

OS

Windows

GPU

Nvidia

CPU

AMD

Ollama version

0.20.2

Originally created by @sdqq1234 on GitHub (Apr 6, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15363 ### What is the issue? I'm using Ollama version 0.20.2 to load a Q4 quantization model of QW3.5 35B. When I connect to the internet and provide a image to query its content, I encounter an infinite "thinking" loop where the content keeps looping without ending. ### Relevant log output ```shell ``` ### OS Windows ### GPU Nvidia ### CPU AMD ### Ollama version 0.20.2
GiteaMirror added the bug label 2026-05-05 02:52:10 -05:00
GiteaMirror commented 2026-05-05 02:52:12 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Apr 6, 2026):

Server logs will aid in debugging.

<!-- gh-comment-id:4191742721 --> @rick-github commented on GitHub (Apr 6, 2026): [Server logs](https://docs.ollama.com/troubleshooting) will aid in debugging.
GiteaMirror commented 2026-05-05 02:52:13 -05:00
Author
Owner
Copy Link

@sdqq1234 commented on GitHub (Apr 6, 2026):

server.log
server-1.log
server-2.log
server-3.log
server-4.log
server-5.log
upgrade.log

I don't know which one is correct, so I uploaded all the server log files.

<!-- gh-comment-id:4192486302 --> @sdqq1234 commented on GitHub (Apr 6, 2026): [server.log](https://github.com/user-attachments/files/26509231/server.log) [server-1.log](https://github.com/user-attachments/files/26509234/server-1.log) [server-2.log](https://github.com/user-attachments/files/26509233/server-2.log) [server-3.log](https://github.com/user-attachments/files/26509232/server-3.log) [server-4.log](https://github.com/user-attachments/files/26509237/server-4.log) [server-5.log](https://github.com/user-attachments/files/26509236/server-5.log) [upgrade.log](https://github.com/user-attachments/files/26509235/upgrade.log) I don't know which one is correct, so I uploaded all the server log files.
GiteaMirror commented 2026-05-05 02:52:14 -05:00
Author
Owner
Copy Link

@pdevine commented on GitHub (Apr 6, 2026):

Which model are you using exactly? If you are using the coding model for non-coding tasks it can get into a loop because the presence_penalty isn't set.

<!-- gh-comment-id:4194277935 --> @pdevine commented on GitHub (Apr 6, 2026): Which model are you using exactly? If you are using the coding model for non-coding tasks it can get into a loop because the presence_penalty isn't set.
GiteaMirror commented 2026-05-05 02:52:15 -05:00
Author
Owner
Copy Link

@andrenaP commented on GitHub (Apr 13, 2026):

Having the same issue using qwen3.5:latest for code generation. I checked it with presence_penalty set to 0 and presence_penalty set to 1.5 it will not exit the thinking mode no matter what presence_penalty I set. If I turn thinking mode off I will get out a good result.

This happens only if I provide a lot of context to the model. It works just fine on simple questions using thinking.

<!-- gh-comment-id:4235408948 --> @andrenaP commented on GitHub (Apr 13, 2026): Having the same issue using qwen3.5:latest for code generation. I checked it with presence_penalty set to 0 and presence_penalty set to 1.5 it will not exit the thinking mode no matter what presence_penalty I set. If I turn thinking mode off I will get out a good result. This happens only if I provide a lot of context to the model. It works just fine on simple questions using thinking.
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#71890
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?