github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 17:49:24 -05:00
Code Issues 14.6k Packages Projects Releases 122 Wiki Activity

[GH-ISSUE #14622] [Bug] lfm2.5-thinking:latest — thinking mode cannot be disabled; #71535

New Issue
Closed
opened 2026-05-05 02:04:51 -05:00 by GiteaMirror · 1 comment
GiteaMirror commented 2026-05-05 02:04:51 -05:00
Owner
Copy Link

Originally created by @Asjad22 on GitHub (Mar 4, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/14622

Bug : Thinking mode cannot be disabled

There is no way to suppress the <think> block output in lfm2.5-thinking:latest.

Expected: When Ollama parameter think is set to false, the model should not output <think> blocks.

Actual: The model always emits a verbose <think> block on every response. think: 'false' isn't working.

Request: Please add support to disable thinking mode via a model parameter (e.g. think: false in Ollama OPTIONS), or provide a non-thinking variant of the model on Ollama.

Relevant log output

curl http://localhost:11434/api/chat \
  -H "Content-Type: application/json" \
  -d '{
    "model": "lfm2.5-thinking:latest",
    "stream": false,
    "options": { "think": false },
    "messages": [
      { "role": "user", "content": "What is 2 + 2?" }
    ]
  }'
{"model":"lfm2.5-thinking:latest","created_at":"2026-03-04T20:18:52.319383911Z","message":{"role":"assistant","content":"The sum of 2 and 2 is simply $2 + 2 = 4$. \n\n\\boxed{4}","thinking":"Okay, let me see. The user asked \"What is 2 + 2?\" Hmm, that seems straightforward, but maybe they want more than just a number. Wait, maybe they're testing if I know basic math? Or perhaps there's a trick here. Let me think.\n\nFirst, 2 plus 2 is a simple arithmetic problem. The answer should be 4. But maybe they expect some deeper answer? Like, in some contexts, 2+2 could symbolize something else? Or maybe it's a play on words? Like \"two plus two\" could be part of a larger question? Wait, the question is just \"What is 2 + 2?\" so probably just 4. \n\nBut let me make sure. If I consider possible contexts: in math class, adding two 2s gives 4. In everyday language, same thing. Unless there's a pun or something. Wait, maybe the user is expecting a different answer? Like, if it's a riddle? For example, \"2 + 2\" could be part of a sentence where it leads to something else. But the question is straightforward. \n\nAlternatively, maybe they want the explanation. Since the user just asked \"What is 2 + 2?\" perhaps they want the answer boxed as 4. Since the instruction says to put the final answer within \\boxed{}, so I should just give 4 in a box. \n\nWait, but maybe I should confirm. Let me check again. If someone asks \"What is 2 + 2?\" the answer is 4. Unless there's a trick, but I can't think of any. So I'll go with 4. Alright, I'll put \\boxed{4}."},"done":true,"done_reason":"stop","total_duration":19064370371,"load_duration":442441364,"prompt_eval_count":17,"prompt_eval_duration":216339639,"eval_count":401,"eval_duration":18312636992}

OS

NAME="Fedora Linux"
VERSION="43 (Workstation Edition)"

GPU

No response

CPU

Model name: Intel(R) Core(TM) i5-10210U CPU @ 1.60GHz

Ollama version

ollama version is 0.14.3

Originally created by @Asjad22 on GitHub (Mar 4, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/14622 ## Bug : Thinking mode cannot be disabled There is no way to suppress the `<think>` block output in `lfm2.5-thinking:latest`. **Expected:** When Ollama parameter `think` is set to `false`, the model should not output `<think>` blocks. **Actual:** The model always emits a verbose `<think>` block on every response. think: 'false' isn't working. **Request:** Please add support to disable thinking mode via a model parameter (e.g. `think: false` in Ollama OPTIONS), or provide a non-thinking variant of the model on Ollama. ### Relevant log output ```shell curl http://localhost:11434/api/chat \ -H "Content-Type: application/json" \ -d '{ "model": "lfm2.5-thinking:latest", "stream": false, "options": { "think": false }, "messages": [ { "role": "user", "content": "What is 2 + 2?" } ] }' {"model":"lfm2.5-thinking:latest","created_at":"2026-03-04T20:18:52.319383911Z","message":{"role":"assistant","content":"The sum of 2 and 2 is simply $2 + 2 = 4$. \n\n\\boxed{4}","thinking":"Okay, let me see. The user asked \"What is 2 + 2?\" Hmm, that seems straightforward, but maybe they want more than just a number. Wait, maybe they're testing if I know basic math? Or perhaps there's a trick here. Let me think.\n\nFirst, 2 plus 2 is a simple arithmetic problem. The answer should be 4. But maybe they expect some deeper answer? Like, in some contexts, 2+2 could symbolize something else? Or maybe it's a play on words? Like \"two plus two\" could be part of a larger question? Wait, the question is just \"What is 2 + 2?\" so probably just 4. \n\nBut let me make sure. If I consider possible contexts: in math class, adding two 2s gives 4. In everyday language, same thing. Unless there's a pun or something. Wait, maybe the user is expecting a different answer? Like, if it's a riddle? For example, \"2 + 2\" could be part of a sentence where it leads to something else. But the question is straightforward. \n\nAlternatively, maybe they want the explanation. Since the user just asked \"What is 2 + 2?\" perhaps they want the answer boxed as 4. Since the instruction says to put the final answer within \\boxed{}, so I should just give 4 in a box. \n\nWait, but maybe I should confirm. Let me check again. If someone asks \"What is 2 + 2?\" the answer is 4. Unless there's a trick, but I can't think of any. So I'll go with 4. Alright, I'll put \\boxed{4}."},"done":true,"done_reason":"stop","total_duration":19064370371,"load_duration":442441364,"prompt_eval_count":17,"prompt_eval_duration":216339639,"eval_count":401,"eval_duration":18312636992} ``` ### OS NAME="Fedora Linux" VERSION="43 (Workstation Edition)" ### GPU _No response_ ### CPU Model name: Intel(R) Core(TM) i5-10210U CPU @ 1.60GHz ### Ollama version ollama version is 0.14.3
GiteaMirror added the bug label 2026-05-05 02:04:51 -05:00
GiteaMirror commented 2026-05-05 02:04:53 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Mar 4, 2026):

The model doesn't have a non-thinking variant, and the template supplied by the model authors shows no mechanism for controlling thinking. It could probably be faked by adding a template that pre-fills the output with an empty <think></think> block, but that's not always reliable.

<!-- gh-comment-id:4000259296 --> @rick-github commented on GitHub (Mar 4, 2026): The model doesn't have a non-thinking variant, and the [template](https://huggingface.co/LiquidAI/LFM2-24B-A2B/blob/main/chat_template.jinja) supplied by the model authors shows no mechanism for controlling thinking. It could probably be faked by adding a template that pre-fills the output with an empty `<think></think>` block, but that's not always reliable.
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#71535
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?