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 #14029] Closed #9169

New Issue
Closed
opened 2026-04-12 22:00:52 -05:00 by GiteaMirror · 11 comments
GiteaMirror commented 2026-04-12 22:00:52 -05:00
Owner
Copy Link

Originally created by @techie-learner on GitHub (Feb 2, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14029

Resolved

Originally created by @techie-learner on GitHub (Feb 2, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14029 Resolved
GiteaMirror added the appfeature request labels 2026-04-12 22:00:52 -05:00
GiteaMirror commented 2026-04-12 22:00:52 -05:00
Author
Owner
Copy Link

@ryanmon1 commented on GitHub (Feb 2, 2026):

Can't you use /set nothink from the CLI to fix this?

<!-- gh-comment-id:3837208940 --> @ryanmon1 commented on GitHub (Feb 2, 2026): Can't you use /set nothink from the CLI to fix this?
GiteaMirror commented 2026-04-12 22:00:53 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 2, 2026):

Can't you use /set nothink from the CLI to fix this?

No. That only affects API calls made from the ollama CLI client, what the OP wants is the ability to control think from the ollama GUI client. An alternative is to use a different client, for example OpenWebUI has a browser-based GUI and the ability to set the value of think.

<!-- gh-comment-id:3837489485 --> @rick-github commented on GitHub (Feb 2, 2026): > Can't you use /set nothink from the CLI to fix this? No. That only affects API calls made from the ollama CLI client, what the OP wants is the ability to control `think` from the ollama GUI client. An alternative is to use a different client, for example OpenWebUI has a browser-based GUI and the ability to set the value of `think`.
GiteaMirror commented 2026-04-12 22:00:53 -05:00
Author
Owner
Copy Link

@techie-learner commented on GitHub (Feb 3, 2026):

Can't you use /set nothink from the CLI to fix this?

Well I will use Ollama GUI app and that why I want that "Thinking" button in Ollama GUI to disable or enable. Ah wait I tried /set nothink even on CMD one and Ollama GUI , and it didn't not worked. It just keeps thinking, so that why I requested a "Thinking" button for ollama GUI.

<!-- gh-comment-id:3839066767 --> @techie-learner commented on GitHub (Feb 3, 2026): > Can't you use /set nothink from the CLI to fix this? Well I will use Ollama GUI app and that why I want that "Thinking" button in Ollama GUI to disable or enable. Ah wait I tried /set nothink even on CMD one and Ollama GUI , and it didn't not worked. It just keeps thinking, so that why I requested a "Thinking" button for ollama GUI.
GiteaMirror commented 2026-04-12 22:00:53 -05:00
Author
Owner
Copy Link

@Nikorasu commented on GitHub (Mar 24, 2026):

I second. Would be nice to have a button somewhere, in that default Ollama GUI, which I actually end up using a lot. (Course the GUI could also use better editing and chat management, but that's not as important as simply being able to disable thinking.)

Alternatively, doing it at the modelfile level would be acceptable. Being able to disable thinking, then /save that to a new modelfile, so we can just use that modelfile in the GUI or other purposes, with thinking disabled by default, would be even more useful.

<!-- gh-comment-id:4119362329 --> @Nikorasu commented on GitHub (Mar 24, 2026): I second. Would be nice to have a button somewhere, in that default Ollama GUI, which I actually end up using a lot. (Course the GUI could also use better editing and chat management, but that's not as important as simply being able to disable thinking.) Alternatively, doing it at the modelfile level would be acceptable. Being able to disable thinking, then /save that to a new modelfile, so we can just use that modelfile in the GUI or other purposes, with thinking disabled by default, would be even more useful.
GiteaMirror commented 2026-04-12 22:00:53 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Mar 24, 2026):

The GUI supports thinking control, but it currently only works for deepseek-v3.1 models.

Image

#10961 for controlling it at the Modelfile level.

<!-- gh-comment-id:4119781023 --> @rick-github commented on GitHub (Mar 24, 2026): The GUI supports thinking control, but it currently only works for deepseek-v3.1 models. <img width="483" height="142" alt="Image" src="https://github.com/user-attachments/assets/04e66826-e87a-4108-aea4-a3599c93c2a6" /> #10961 for controlling it at the Modelfile level.
GiteaMirror commented 2026-04-12 22:00:54 -05:00
Author
Owner
Copy Link

@forbiddenPHP commented on GitHub (Apr 2, 2026):

BUMP! It's a pain in the a to have loops of thinking from whatever model! PLEASE hear us! <3

<!-- gh-comment-id:4174270343 --> @forbiddenPHP commented on GitHub (Apr 2, 2026): BUMP! It's a pain in the a to have loops of thinking from whatever model! PLEASE hear us! <3
GiteaMirror commented 2026-04-12 22:00:54 -05:00
Author
Owner
Copy Link

@Rastafabisch commented on GitHub (Apr 2, 2026):

This would really useful for the QWEN 3.5 models! :)

<!-- gh-comment-id:4176146513 --> @Rastafabisch commented on GitHub (Apr 2, 2026): This would really useful for the QWEN 3.5 models! :)
GiteaMirror commented 2026-04-12 22:00:54 -05:00
Author
Owner
Copy Link

@dhirajlochib commented on GitHub (Apr 2, 2026):

Implemented in PR #15217.

Two-file frontend change:

  • useModelCapabilities.ts: added useHasThinkingCapability hook (mirrors the existing useHasVisionCapability / useHasToolsCapability pattern — checks for 'thinking' in the capabilities API response)
  • ChatForm.tsx: replaced selectedModel?.model.toLowerCase().startsWith('deepseek-v3.1') with useHasThinkingCapability(selectedModel?.model)

The modelSupportsThinkingLevels check (gpt-oss low/medium/high) is left as-is since that's a separate product feature.

<!-- gh-comment-id:4178494154 --> @dhirajlochib commented on GitHub (Apr 2, 2026): Implemented in PR #15217. Two-file frontend change: - `useModelCapabilities.ts`: added `useHasThinkingCapability` hook (mirrors the existing `useHasVisionCapability` / `useHasToolsCapability` pattern — checks for `'thinking'` in the capabilities API response) - `ChatForm.tsx`: replaced `selectedModel?.model.toLowerCase().startsWith('deepseek-v3.1')` with `useHasThinkingCapability(selectedModel?.model)` The `modelSupportsThinkingLevels` check (gpt-oss low/medium/high) is left as-is since that's a separate product feature.
GiteaMirror commented 2026-04-12 22:00:54 -05:00
Author
Owner
Copy Link

@zamiere commented on GitHub (Apr 3, 2026):

Creating a fake deepseek-v3.1-something model doesn’t solve the problem with the gemma4 model. The button appeared, but it doesn’t work. (0.9.0 ollama version)

<!-- gh-comment-id:4183578384 --> @zamiere commented on GitHub (Apr 3, 2026): Creating a fake deepseek-v3.1-something model doesn’t solve the problem with the gemma4 model. The button appeared, but it doesn’t work. (0.9.0 ollama version)
GiteaMirror commented 2026-04-12 22:00:55 -05:00
Author
Owner
Copy Link

@techie-learner commented on GitHub (Apr 7, 2026):

Creating a fake deepseek-v3.1-something model doesn’t solve the problem with the gemma4 model. The button appeared, but it doesn’t work. (0.9.0 ollama version)

Obviously. Renaming a model doesn't rewrite its architecture. You just tricked the UI into showing a button that the backend doesn't know how to use for Gemma yet. We are waiting for an actual code implementation, not a nickname change. So if you change a car named "Buggati"(Gemma4) to "Lamborghini"(deepseek-v3.1-something model name) and painted it in "Lamborghini" style, the person(The thinking toggle button) who knowns to use Lamborghini will come to use because he is tricked but he will always failed at fake lamborghini style buggati as they are different to each other. Basic common sense.

<!-- gh-comment-id:4196991794 --> @techie-learner commented on GitHub (Apr 7, 2026): > Creating a fake deepseek-v3.1-something model doesn’t solve the problem with the gemma4 model. The button appeared, but it doesn’t work. (0.9.0 ollama version) Obviously. Renaming a model doesn't rewrite its architecture. You just tricked the UI into showing a button that the backend doesn't know how to use for Gemma yet. We are waiting for an actual code implementation, not a nickname change. So if you change a car named "Buggati"(Gemma4) to "Lamborghini"(deepseek-v3.1-something model name) and painted it in "Lamborghini" style, the person(The thinking toggle button) who knowns to use Lamborghini will come to use because he is tricked but he will always failed at fake lamborghini style buggati as they are different to each other. Basic common sense.
GiteaMirror commented 2026-04-12 22:00:55 -05:00
Author
Owner
Copy Link

@techie-learner commented on GitHub (Apr 7, 2026):

This would really useful for the QWEN 3.5 models! :)

The logic we're discussing covers all models with thinking architectures. Once the PR is merged, it won't matter if it's Qwen, Gemma, or DeepSeek—the toggle will work for everything capable of reasoning. We're fixing the engine, not just one engine's piston that only support one model type.

<!-- gh-comment-id:4197020976 --> @techie-learner commented on GitHub (Apr 7, 2026): > This would really useful for the QWEN 3.5 models! :) The logic we're discussing covers all models with thinking architectures. Once the PR is merged, it won't matter if it's Qwen, Gemma, or DeepSeek—the toggle will work for everything capable of reasoning. We're fixing the engine, not just one engine's piston that only support one model type.
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 app feature request
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#9169
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?