[GH-ISSUE #15857] registry.ollama.ai/library/qwen3.6-35b-a3b-tool:latest does not support tools #72165

Open
opened 2026-05-05 03:34:56 -05:00 by GiteaMirror · 4 comments
Owner

Originally created by @tanmingli on GitHub (Apr 28, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15857

registry.ollama.ai/library/qwen3.6-35b-a3b-tool:latest does not support tools

Originally created by @tanmingli on GitHub (Apr 28, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15857 registry.ollama.ai/library/qwen3.6-35b-a3b-tool:latest does not support tools
Author
Owner

@noochboy commented on GitHub (Apr 28, 2026):

Hi,

I've been having issues with tool calls in general since a couple of releases.

It's not longer working with qwen3.5, gemma4 neither,

local models - not working
NAME ID SIZE MODIFIED
qwen3.5:latest 6488c96fa5fa 6.6 GB 9 days ago

tool calling for the cloud models works

gemma4:31b-cloud c382fbfbc73b - 11 days ago
glm-5.1:cloud 59472abf9d0a - 12 days ago
minimax-m2.7:cloud 06daa293c105 - 2 weeks ago

Symptom: It just pasts what it wants to do, but doesn't do anything in the end

check the TASKS.md file, you should find info there

<reply - from hermes agent>
I'll read the TASKS.md file to see what information it contains.

 bash
 cat TASKS.md
  • something seems wrong in the logic. It keeps on going around in circles

Gemma4:

can you list the content of file /home/noochboy/.hermes/profiles/researcher/FIELDS.md

<reply - from hermes agent>
I do not have the ability to browse the local file system for arbitrary files like /home/noochboy/profiles/files/noochboy_profile/FIELDS.md.

 If this file is content you provided in the chat history, please paste it here, and I will be happy to read, summarize, or analyze it for you. Otherwise, I cannot access it.
<!-- gh-comment-id:4333678239 --> @noochboy commented on GitHub (Apr 28, 2026): Hi, I've been having issues with tool calls in general since a couple of releases. It's not longer working with qwen3.5, gemma4 neither, local models - not working NAME ID SIZE MODIFIED qwen3.5:latest 6488c96fa5fa 6.6 GB 9 days ago tool calling for the cloud models works gemma4:31b-cloud c382fbfbc73b - 11 days ago glm-5.1:cloud 59472abf9d0a - 12 days ago minimax-m2.7:cloud 06daa293c105 - 2 weeks ago Symptom: It just pasts what it wants to do, but doesn't do anything in the end <prompt> check the TASKS.md file, you should find info there </prompt> <reply - from hermes agent> I'll read the TASKS.md file to see what information it contains. bash cat TASKS.md </reply> + something seems wrong in the logic. It keeps on going around in circles Gemma4: <prompt> can you list the content of file /home/noochboy/.hermes/profiles/researcher/FIELDS.md </prompt> <reply - from hermes agent> I do not have the ability to browse the local file system for arbitrary files like /home/noochboy/profiles/files/noochboy_profile/FIELDS.md. If this file is content you provided in the chat history, please paste it here, and I will be happy to read, summarize, or analyze it for you. Otherwise, I cannot access it. </reply>
Author
Owner

@tanmingli commented on GitHub (Apr 30, 2026):

ollama show qwen3.6-35b-a3b-tool:latest
Model
architecture qwen35moe
parameters 34.7B
context length 262144
embedding length 2048
quantization Q4_K_M

Capabilities
completion

Parameters
num_ctx 65536
num_gpu 99
stop "<|im_end|>"
stop "<|endoftext|>"
stop "<|tool_call_end|>"
temperature 0.7
top_p 0.9

System
You are a helpful assistant.

<!-- gh-comment-id:4350157715 --> @tanmingli commented on GitHub (Apr 30, 2026): ollama show qwen3.6-35b-a3b-tool:latest Model architecture qwen35moe parameters 34.7B context length 262144 embedding length 2048 quantization Q4_K_M Capabilities completion Parameters num_ctx 65536 num_gpu 99 stop "<|im_end|>" stop "<|endoftext|>" stop "<|tool_call_end|>" temperature 0.7 top_p 0.9 System You are a helpful assistant.
Author
Owner

@tanmingli commented on GitHub (Apr 30, 2026):

Help check if my configuration is correct

<!-- gh-comment-id:4350160217 --> @tanmingli commented on GitHub (Apr 30, 2026): Help check if my configuration is correct
Author
Owner

@ArkaD171717 commented on GitHub (May 1, 2026):

Root cause: qwen35moe isn't in the architecture auto-detection switch in server/create.go (lines 525-547), so parser/renderer stay empty and HasToolSupport() is never called. ollama show confirms only completion capability. Workaround: ollama create qwen3.6-tool-fixed --from qwen3.6-35b-a3b-tool --parser qwen3.5 --renderer qwen3.5.
I have a reference fix at #15899 if it helps

<!-- gh-comment-id:4359001901 --> @ArkaD171717 commented on GitHub (May 1, 2026): Root cause: qwen35moe isn't in the architecture auto-detection switch in server/create.go (lines 525-547), so parser/renderer stay empty and HasToolSupport() is never called. ollama show confirms only completion capability. Workaround: ollama create qwen3.6-tool-fixed --from qwen3.6-35b-a3b-tool --parser qwen3.5 --renderer qwen3.5. I have a reference fix at #15899 if it helps
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#72165