github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #12589] gpt-oss:120b no longer obeying Reasoning Effort setting #8354

New Issue
Closed
opened 2026-04-12 20:57:17 -05:00 by GiteaMirror · 2 comments
GiteaMirror commented 2026-04-12 20:57:17 -05:00
Owner
Copy Link

Originally created by @asitwere on GitHub (Oct 12, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12589

What is the issue?

Following this week's update of GPT-OSS:120B, low/medium/high "Reasoning Effort" settings are no longer honored.

STEPS:

  • Downloaded new model
  • Exported Modelfile
  • Revised "Reasoning: medium" to "Reasoning: low" & "Reasoning: high" in new Modelfiles
  • Created new custom models for "low" and "high"

RESULT:

  • Reasoning effort is consistently the same across the standard model ("medium") and custom models ("low" and "high")

Testing with pre-update prompts:

  • I tested with previously-run prompts for "low" and "high" which had yielded 2 seconds to 4 minutes of reasoning (i.e. "low" to "high") under the previous model release
  • The new model release consistently output 20-30 seconds of reasoning (i.e. "medium") across all Modelfiles, regardless of Reasoning setting

Additional attempt in Open WebUI:

  • I also attempted to adjust "Reasoning Effort" parameter of the standard model to "low/high" in Open WebUI and the model honored "medium" reasoning length for each prompt, regardless of "Reasoning Effort" setting

(Working with an M3 Mac Studio running Ollama 0.12.5)

Relevant log output

OS

macOS

GPU

Apple

CPU

Apple

Ollama version

0.12.5

Originally created by @asitwere on GitHub (Oct 12, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12589 ### What is the issue? Following this week's update of GPT-OSS:120B, low/medium/high "Reasoning Effort" settings are no longer honored. ### STEPS: - Downloaded new model - Exported Modelfile - Revised "Reasoning: medium" to "Reasoning: low" & "Reasoning: high" in new Modelfiles - Created new custom models for "low" and "high" ### RESULT: - Reasoning effort is consistently the same across the standard model ("medium") and custom models ("low" and "high") Testing with pre-update prompts: - I tested with previously-run prompts for "low" and "high" which had yielded **2 seconds to 4 minutes of reasoning (i.e. "low" to "high")** under the previous model release - The new model release consistently output **20-30 seconds of reasoning (i.e. "medium")** across all Modelfiles, regardless of Reasoning setting Additional attempt in Open WebUI: - I also attempted to adjust "Reasoning Effort" parameter of the standard model to "low/high" in Open WebUI and the model honored "medium" reasoning length for each prompt, regardless of "Reasoning Effort" setting (Working with an M3 Mac Studio running Ollama 0.12.5) ### Relevant log output ```shell ``` ### OS macOS ### GPU Apple ### CPU Apple ### Ollama version 0.12.5
GiteaMirror added the bug label 2026-04-12 20:57:17 -05:00
GiteaMirror commented 2026-04-12 20:57:19 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Oct 12, 2025):

Ollama now enables thinking in all thinking-capable models (#12533). Since your inference call is not explicitly setting the reasoning effort, it gets set to "medium", which overrides the default in the Modelfile. If you want to modify the template to control thinking, you need to remove the lines that use the value of the think field.

--- Modelfile.orig      2025-10-12 23:30:33.975381079 +0200
+++ Modelfile.low       2025-10-12 23:29:23.459228412 +0200
@@ -5,13 +5,8 @@
 TEMPLATE """<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI.
 Knowledge cutoff: 2024-06
 Current date: {{ currentDate }}
-{{- if and .IsThinkSet .Think (ne .ThinkLevel "") }}
 
-Reasoning: {{ .ThinkLevel }}
-{{- else if or (not .IsThinkSet) (and .IsThinkSet .Think) }}
-
-Reasoning: medium
-{{- end }}
+Reasoning: low
 
 {{- $hasNonBuiltinTools := false }}
 {{- if .Tools -}}
<!-- gh-comment-id:3395382495 --> @rick-github commented on GitHub (Oct 12, 2025): Ollama now enables thinking in all thinking-capable models (#12533). Since your inference call is not explicitly setting the reasoning effort, it gets set to "medium", which overrides the default in the Modelfile. If you want to modify the template to control thinking, you need to remove the lines that use the value of the `think` field. ```diff --- Modelfile.orig 2025-10-12 23:30:33.975381079 +0200 +++ Modelfile.low 2025-10-12 23:29:23.459228412 +0200 @@ -5,13 +5,8 @@ TEMPLATE """<|start|>system<|message|>You are ChatGPT, a large language model trained by OpenAI. Knowledge cutoff: 2024-06 Current date: {{ currentDate }} -{{- if and .IsThinkSet .Think (ne .ThinkLevel "") }} -Reasoning: {{ .ThinkLevel }} -{{- else if or (not .IsThinkSet) (and .IsThinkSet .Think) }} - -Reasoning: medium -{{- end }} +Reasoning: low {{- $hasNonBuiltinTools := false }} {{- if .Tools -}} ```
GiteaMirror commented 2026-04-12 20:57:20 -05:00
Author
Owner
Copy Link

@asitwere commented on GitHub (Oct 12, 2025):

Thanks, new Modelfiles were successful

<!-- gh-comment-id:3395419844 --> @asitwere commented on GitHub (Oct 12, 2025): Thanks, new Modelfiles were successful
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#8354
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?