github-starred/open-webui
2
0
Fork 0
mirror of https://github.com/open-webui/open-webui.git synced 2026-05-07 03:18:23 -05:00
Code Issues 1.3k Packages Projects Releases 126 Wiki Activity

[GH-ISSUE #16744] issue: Harmony Format Support Issue with gpt-oss Models via LM Studio #56695

New Issue
Closed
opened 2026-05-05 19:55:57 -05:00 by GiteaMirror · 8 comments
GiteaMirror commented 2026-05-05 19:55:57 -05:00
Owner
Copy Link

Originally created by @Pixellevel on GitHub (Aug 20, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/16744

Check Existing Issues

  • I have searched the existing issues and discussions.
  • I am using the latest version of Open WebUI.

Installation Method

Pip Install

Open WebUI Version

v0.6.22

Ollama Version (if applicable)

No response

Operating System

Windows 10

Browser (if applicable)

Chrome 128.0.6613.85

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 included the browser console logs.
  • I have included the Docker container logs.
  • I have provided every relevant configuration, setting, and environment variable used in my setup.
  • I have clearly listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc).
  • I have documented step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation. My steps:
  • Start with the initial platform/version/OS and dependencies used,
  • Specify exact install/launch/configure commands,
  • List URLs visited, user input (incl. example values/emails/passwords if needed),
  • Describe all options and toggles enabled or changed,
  • Include any files or environmental changes,
  • Identify the expected and actual result at each stage,
  • Ensure any reasonably skilled user can follow and hit the same issue.

Expected Behavior

When using gpt-oss models gpt-oss-20b through LM Studio, adding "Reasoning: high" to the system prompt in Open WebUI should:
Activate the model's high-level reasoning mode
Display detailed chain-of-thought processing in responses
Show enhanced analytical capabilities as designed by OpenAI's Harmony format
The gpt-oss models were specifically trained on OpenAI's Harmony response format and support configurable reasoning levels (low, medium, high) through system prompt directives.

Actual Behavior

When Reasoning: high is added to the system prompt in Open WebUI:
The reasoning directive appears to be ignored
The model responds in standard mode without enhanced reasoning
No chain-of-thought processing is visible
The response quality/depth doesn't reflect the requested reasoning level

Steps to Reproduce

1.Setup LM Studio:
Download and install LM Studio
Download gpt-oss-20b model from OpenAI via LM Studio's model manager
Start LM Studio server with default settings (typically localhost:1234)
Verify the model loads correctly and responds to basic queries
2.Configure Open WebUI:
install Open WebUI
Navigate to the Administrator Panel, Settings, External Connections,
Add OpenAI API connectionhttp://127.0.0.1:1234/v1 Key: NONE
Verify connection shows as active
3.Test Standard Behavior:
Create new chat in Open WebUI
Select the gpt-oss-20b model
Send a complex query (e.g., "Explain quantum entanglement and its applications")
Note the response style and depth
4.Test with Reasoning Directive:
Click Advanced Dialogue Settings, and the system prompts you to enter the word: Reasoning: high
SAlternatively, select the Administrator Panel, then Settings, then Model, then select gbt-oss-20b and add the prompt Reasoning: high and save.
Create new chat with same model
Send the same complex query
Compare response - should show enhanced reasoning but doesn't
5.Verify LM Studio Direct Access:
Test the same prompt directly in LM Studio interface
You can directly select the reasoning strength below the l m studio chat window
Confirm if reasoning enhancement works when bypassing Open WebUI

Logs & Screenshots

Image Image Image

Additional Information

No response

Originally created by @Pixellevel on GitHub (Aug 20, 2025). Original GitHub issue: https://github.com/open-webui/open-webui/issues/16744 ### Check Existing Issues - [x] I have searched the existing issues and discussions. - [x] I am using the latest version of Open WebUI. ### Installation Method Pip Install ### Open WebUI Version v0.6.22 ### Ollama Version (if applicable) _No response_ ### Operating System Windows 10 ### Browser (if applicable) Chrome 128.0.6613.85 ### 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 included the browser console logs. - [x] I have included the Docker container logs. - [x] I have **provided every relevant configuration, setting, and environment variable used in my setup.** - [x] I have clearly **listed every relevant configuration, custom setting, environment variable, and command-line option that influences my setup** (such as Docker Compose overrides, .env values, browser settings, authentication configurations, etc). - [x] I have documented **step-by-step reproduction instructions that are precise, sequential, and leave nothing to interpretation**. My steps: - Start with the initial platform/version/OS and dependencies used, - Specify exact install/launch/configure commands, - List URLs visited, user input (incl. example values/emails/passwords if needed), - Describe all options and toggles enabled or changed, - Include any files or environmental changes, - Identify the expected and actual result at each stage, - Ensure any reasonably skilled user can follow and hit the same issue. ### Expected Behavior When using gpt-oss models gpt-oss-20b through LM Studio, adding "Reasoning: high" to the system prompt in Open WebUI should: Activate the model's high-level reasoning mode Display detailed chain-of-thought processing in responses Show enhanced analytical capabilities as designed by OpenAI's Harmony format The gpt-oss models were specifically trained on OpenAI's Harmony response format and support configurable reasoning levels (low, medium, high) through system prompt directives. ### Actual Behavior When Reasoning: high is added to the system prompt in Open WebUI: The reasoning directive appears to be ignored The model responds in standard mode without enhanced reasoning No chain-of-thought processing is visible The response quality/depth doesn't reflect the requested reasoning level ### Steps to Reproduce 1.Setup LM Studio: Download and install LM Studio Download gpt-oss-20b model from OpenAI via LM Studio's model manager Start LM Studio server with default settings (typically localhost:1234) Verify the model loads correctly and responds to basic queries 2.Configure Open WebUI: install Open WebUI Navigate to the Administrator Panel, Settings, External Connections, Add OpenAI API connectionhttp://127.0.0.1:1234/v1 Key: NONE Verify connection shows as active 3.Test Standard Behavior: Create new chat in Open WebUI Select the gpt-oss-20b model Send a complex query (e.g., "Explain quantum entanglement and its applications") Note the response style and depth 4.Test with Reasoning Directive: Click Advanced Dialogue Settings, and the system prompts you to enter the word: Reasoning: high SAlternatively, select the Administrator Panel, then Settings, then Model, then select gbt-oss-20b and add the prompt Reasoning: high and save. Create new chat with same model Send the same complex query Compare response - should show enhanced reasoning but doesn't 5.Verify LM Studio Direct Access: Test the same prompt directly in LM Studio interface You can directly select the reasoning strength below the l m studio chat window Confirm if reasoning enhancement works when bypassing Open WebUI ### Logs & Screenshots <img width="1445" height="925" alt="Image" src="https://github.com/user-attachments/assets/8b83b151-904f-4ed6-b456-4a690fbc9728" /> <img width="1641" height="651" alt="Image" src="https://github.com/user-attachments/assets/3bebefbe-fe46-4c3e-b7fd-b4acc1005b4d" /> <img width="1478" height="790" alt="Image" src="https://github.com/user-attachments/assets/76cfd46a-b6fc-4537-825b-e4a677d886cf" /> ### Additional Information _No response_
GiteaMirror added the bug label 2026-05-05 19:55:58 -05:00
GiteaMirror commented 2026-05-05 19:56:09 -05:00
Author
Owner
Copy Link

@Pixellevel commented on GitHub (Aug 20, 2025):

Root Cause Hypothesis:
The issue likely stems from Open WebUI not properly preserving or translating the Harmony format requirements when communicating with LM Studio. While LM Studio supports Harmony format natively, Open WebUI may be using standard OpenAI API formatting that strips or doesn't properly embed the Harmony-specific directives.
Request:
Could Open WebUI add native support for Harmony format system directives, particularly for reasoning levels (Reasoning: low/medium/high) when interfacing with gpt-oss models? This would ensure seamless integration with OpenAI's open-weight model series.
Additional Context:
According to OpenAI's documentation, gpt-oss models were specifically trained on the Harmony response format and should only be used with this format for optimal performance.

<!-- gh-comment-id:3204852821 --> @Pixellevel commented on GitHub (Aug 20, 2025): Root Cause Hypothesis: The issue likely stems from Open WebUI not properly preserving or translating the Harmony format requirements when communicating with LM Studio. While LM Studio supports Harmony format natively, Open WebUI may be using standard OpenAI API formatting that strips or doesn't properly embed the Harmony-specific directives. Request: Could Open WebUI add native support for Harmony format system directives, particularly for reasoning levels (Reasoning: low/medium/high) when interfacing with gpt-oss models? This would ensure seamless integration with OpenAI's open-weight model series. Additional Context: According to OpenAI's documentation, gpt-oss models were specifically trained on the Harmony response format and should only be used with this format for optimal performance.
GiteaMirror commented 2026-05-05 19:56:10 -05:00
Author
Owner
Copy Link

@tjbck commented on GitHub (Aug 20, 2025):

You set the system prompt and did not correctly configure the parameter.

<!-- gh-comment-id:3204959060 --> @tjbck commented on GitHub (Aug 20, 2025): You set the system prompt and did not correctly configure the parameter.
GiteaMirror commented 2026-05-05 19:56:13 -05:00
Author
Owner
Copy Link

@Pixellevel commented on GitHub (Aug 20, 2025):

You set the system prompt and did not correctly configure the parameter.

Thank you for the response! Could you please show me exactly where and how to configure the reasoning parameter in Open WebUI? I tried adding Reasoning: high to the system prompt but it seems I'm missing something. What's the correct way to set this up?

<!-- gh-comment-id:3204987943 --> @Pixellevel commented on GitHub (Aug 20, 2025): > You set the system prompt and did not correctly configure the parameter. Thank you for the response! Could you please show me exactly where and how to configure the reasoning parameter in Open WebUI? I tried adding Reasoning: high to the system prompt but it seems I'm missing something. What's the correct way to set this up?
GiteaMirror commented 2026-05-05 19:56:16 -05:00
Author
Owner
Copy Link

@Pixellevel commented on GitHub (Aug 20, 2025):

@tjbck Thank you for the response! Could you please show me exactly where and how to configure the reasoning parameter in Open WebUI? I tried adding Reasoning: high to the system prompt but it seems I'm missing something. What's the correct way to set this up?

<!-- gh-comment-id:3205025355 --> @Pixellevel commented on GitHub (Aug 20, 2025): @tjbck Thank you for the response! Could you please show me exactly where and how to configure the reasoning parameter in Open WebUI? I tried adding Reasoning: high to the system prompt but it seems I'm missing something. What's the correct way to set this up?
GiteaMirror commented 2026-05-05 19:56:20 -05:00
Author
Owner
Copy Link

@Pixellevel commented on GitHub (Aug 20, 2025):

https://huggingface.co/openai/gpt-oss-120b#:~:text=Reasoning%20levels,g.%2C%20%22Reasoning%3A%20high%22.
Because I saw here that he only needs to add this in the system prompt words

<!-- gh-comment-id:3205043337 --> @Pixellevel commented on GitHub (Aug 20, 2025): https://huggingface.co/openai/gpt-oss-120b#:~:text=Reasoning%20levels,g.%2C%20%22Reasoning%3A%20high%22. Because I saw here that he only needs to add this in the system prompt words
GiteaMirror commented 2026-05-05 19:56:21 -05:00
Author
Owner
Copy Link

@Pixellevel commented on GitHub (Aug 20, 2025):

Image I tried the Reasoning Effort parameter, but it didn't work.
<!-- gh-comment-id:3205081393 --> @Pixellevel commented on GitHub (Aug 20, 2025): <img width="555" height="253" alt="Image" src="https://github.com/user-attachments/assets/3536a95d-d059-4d2d-b55c-d2c13e541f3a" /> I tried the Reasoning Effort parameter, but it didn't work.
GiteaMirror commented 2026-05-05 19:56:21 -05:00
Author
Owner
Copy Link

@Pixellevel commented on GitHub (Aug 20, 2025):

@tjbck Could you please explain it to me?

<!-- gh-comment-id:3205162926 --> @Pixellevel commented on GitHub (Aug 20, 2025): @tjbck Could you please explain it to me?
GiteaMirror commented 2026-05-05 19:56:21 -05:00
Author
Owner
Copy Link

@MountainX commented on GitHub (Sep 5, 2025):

I'm facing this issue too and would like to know what is the solution.

<!-- gh-comment-id:3259382409 --> @MountainX commented on GitHub (Sep 5, 2025): I'm facing this issue too and would like to know what is the solution.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
bug
confirmed
confirmed issue
core
documentation
enhancement
good first issue
help wanted
non-core
pull-request
Mirrored from GitHub Pull Request
 python
question
testing wanted
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/open-webui#56695
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?