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 #11112] 0.9.1 breaks tool calling for qwen3 #69390

New Issue
Closed
opened 2026-05-04 17:58:16 -05:00 by GiteaMirror · 3 comments
GiteaMirror commented 2026-05-04 17:58:16 -05:00
Owner
Copy Link

Originally created by @Devansh-bit on GitHub (Jun 18, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/11112

Originally assigned to: @jmorganca on GitHub.

What is the issue?

The model shows its intent to call tools in its thinking, but neither there is any tool call associated with the response nor there is a normal response string. Same code works perfectly with 0.9.0

llm = ChatOllama(model="qwen3:8b", temperature=0.0, seed=1)
self.llm = llm.bind_tools([get_current_datetime])
test_response = self.llm.invoke([
    SystemMessage("You are a helpful assistant. Use tools when needed."),
    HumanMessage("What's the current date and time?")
])
print(f"Test response: {test_response}")

Response in 0.9.1

Test response: content="<think>\nOkay, the user is asking for the current date and time. Let me check the tools available. There's a function called get_current_datetime that returns the datetime in a specific format. I don't need any parameters for this. I'll call that function to get the answer.\n</think>\n\n" additional_kwargs={} response_metadata={'model': 'qwen3:8b', 'created_at': '2025-06-18T11:53:56.7284535Z', 'done': True, 'done_reason': 'stop', 'total_duration': 4991868400, 'load_duration': 3131632000, 'prompt_eval_count': 541, 'prompt_eval_duration': 432141200, 'eval_count': 77, 'eval_duration': 1422343500, 'model_name': 'qwen3:8b'} id='run--c4f3f0a6-e6cb-41e7-ba3c-0453f1d9e491-0' usage_metadata={'input_tokens': 541, 'output_tokens': 77, 'total_tokens': 618}

Previous response (in 0.9.0)

Test response: content="<think>\nOkay, the user is asking for the current date and time. Let me check the tools available. There's a function called get_current_datetime that returns the datetime in a specific format. I don't need any parameters for this. I'll call that function to get the answer.\n</think>\n\n" additional_kwargs={} response_metadata={'model': 'qwen3:8b', 'created_at': '2025-06-18T11:48:57.9530064Z', 'done': True, 'done_reason': 'stop', 'total_duration': 1855447600, 'load_duration': 17265100, 'prompt_eval_count': 541, 'prompt_eval_duration': 505970800, 'eval_count': 77, 'eval_duration': 1327688200, 'model_name': 'qwen3:8b'} id='run--676df2f1-1df0-4ba9-a48d-c81ea99951a3-0' tool_calls=[{'name': 'get_current_datetime', 'args': {}, 'id': 'db693716-c8a7-4383-b85e-0ac6618a631e', 'type': 'tool_call'}] usage_metadata={'input_tokens': 541, 'output_tokens': 77, 'total_tokens': 618}

OS

Windows

GPU

RTX 3070 laptop GPU

CPU

i7-11800H

Ollama version

0.9.1

Originally created by @Devansh-bit on GitHub (Jun 18, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/11112 Originally assigned to: @jmorganca on GitHub. ### What is the issue? The model shows its intent to call tools in its thinking, but neither there is any tool call associated with the response nor there is a normal response string. Same code works perfectly with 0.9.0 ```python llm = ChatOllama(model="qwen3:8b", temperature=0.0, seed=1) self.llm = llm.bind_tools([get_current_datetime]) test_response = self.llm.invoke([ SystemMessage("You are a helpful assistant. Use tools when needed."), HumanMessage("What's the current date and time?") ]) print(f"Test response: {test_response}") ``` Response in 0.9.1 ```shell Test response: content="<think>\nOkay, the user is asking for the current date and time. Let me check the tools available. There's a function called get_current_datetime that returns the datetime in a specific format. I don't need any parameters for this. I'll call that function to get the answer.\n</think>\n\n" additional_kwargs={} response_metadata={'model': 'qwen3:8b', 'created_at': '2025-06-18T11:53:56.7284535Z', 'done': True, 'done_reason': 'stop', 'total_duration': 4991868400, 'load_duration': 3131632000, 'prompt_eval_count': 541, 'prompt_eval_duration': 432141200, 'eval_count': 77, 'eval_duration': 1422343500, 'model_name': 'qwen3:8b'} id='run--c4f3f0a6-e6cb-41e7-ba3c-0453f1d9e491-0' usage_metadata={'input_tokens': 541, 'output_tokens': 77, 'total_tokens': 618} ``` Previous response (in 0.9.0) ```shell Test response: content="<think>\nOkay, the user is asking for the current date and time. Let me check the tools available. There's a function called get_current_datetime that returns the datetime in a specific format. I don't need any parameters for this. I'll call that function to get the answer.\n</think>\n\n" additional_kwargs={} response_metadata={'model': 'qwen3:8b', 'created_at': '2025-06-18T11:48:57.9530064Z', 'done': True, 'done_reason': 'stop', 'total_duration': 1855447600, 'load_duration': 17265100, 'prompt_eval_count': 541, 'prompt_eval_duration': 505970800, 'eval_count': 77, 'eval_duration': 1327688200, 'model_name': 'qwen3:8b'} id='run--676df2f1-1df0-4ba9-a48d-c81ea99951a3-0' tool_calls=[{'name': 'get_current_datetime', 'args': {}, 'id': 'db693716-c8a7-4383-b85e-0ac6618a631e', 'type': 'tool_call'}] usage_metadata={'input_tokens': 541, 'output_tokens': 77, 'total_tokens': 618} ``` ### OS Windows ### GPU RTX 3070 laptop GPU ### CPU i7-11800H ### Ollama version 0.9.1
GiteaMirror added the bug label 2026-05-04 17:58:16 -05:00
GiteaMirror commented 2026-05-04 17:58:21 -05:00
Author
Owner
Copy Link

@jmorganca commented on GitHub (Jun 18, 2025):

Sorry about this. A fix is on the way in https://github.com/ollama/ollama/pull/11113 and will be released in 0.9.2.

<!-- gh-comment-id:2983938347 --> @jmorganca commented on GitHub (Jun 18, 2025): Sorry about this. A fix is on the way in https://github.com/ollama/ollama/pull/11113 and will be released in 0.9.2.
GiteaMirror commented 2026-05-04 17:58:22 -05:00
Author
Owner
Copy Link

@Devansh-bit commented on GitHub (Jun 18, 2025):

That was quick! Thank you

<!-- gh-comment-id:2984141982 --> @Devansh-bit commented on GitHub (Jun 18, 2025): That was quick! Thank you
GiteaMirror commented 2026-05-04 17:58:24 -05:00
Author
Owner
Copy Link

@jmorganca commented on GitHub (Jun 18, 2025):

@Devansh-bit 0.9.2 is out now and should have this fixed. Thanks for reporting!

<!-- gh-comment-id:2984846086 --> @jmorganca commented on GitHub (Jun 18, 2025): @Devansh-bit 0.9.2 is out now and should have this fixed. Thanks for reporting!
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#69390
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?