github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-08 00:51:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #13768] Hugging Face models "does not support tools" #71083

New Issue
Closed
opened 2026-05-04 23:57:33 -05:00 by GiteaMirror · 8 comments
GiteaMirror commented 2026-05-04 23:57:33 -05:00
Owner
Copy Link

Originally created by @Raboo on GitHub (Jan 18, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/13768

What is the issue?

Lets say I fetch a huggingface model like hf.co/mradermacher/MiniMax-M2.1-i1-GGUF:Q4_K_M.
When I try to use it for agentic coding, Ollama says "does not support tools".
The model itself does support tools. But Ollama doesn't know that. How do you go about to tell Ollama that a HF model support tools?

Relevant log output

OS

No response

GPU

No response

CPU

No response

Ollama version

0.14.2

Originally created by @Raboo on GitHub (Jan 18, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/13768 ### What is the issue? Lets say I fetch a huggingface model like `hf.co/mradermacher/MiniMax-M2.1-i1-GGUF:Q4_K_M`. When I try to use it for agentic coding, Ollama says "does not support tools". The model itself does support tools. But Ollama doesn't know that. How do you go about to tell Ollama that a HF model support tools? ### Relevant log output ```shell ``` ### OS _No response_ ### GPU _No response_ ### CPU _No response_ ### Ollama version 0.14.2
GiteaMirror added the bug label 2026-05-04 23:57:33 -05:00
GiteaMirror commented 2026-05-04 23:57:37 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 18, 2026):

The model requires a parser or a template that enables tool use.

<!-- gh-comment-id:3765871362 --> @rick-github commented on GitHub (Jan 18, 2026): The model requires a [parser](https://github.com/ollama/ollama/pull/13582) or a [template](https://ollama.com/frob/minimax-m2.1:latest/blobs/6dc1fe3f1a44) that enables tool use.
GiteaMirror commented 2026-05-04 23:57:38 -05:00
Author
Owner
Copy Link

@Raboo commented on GitHub (Jan 19, 2026):

But if I haven't imported the model to ollama registry, how do I achieve that when loading models directly from HF?
Do I create a model file with?

FROM hf.co/mradermacher/MiniMax-M2.1-i1-GGUF:Q4_K_M

The model exists now in the registry and I can use that. But for future models..

<!-- gh-comment-id:3765962118 --> @Raboo commented on GitHub (Jan 19, 2026): But if I haven't imported the model to ollama registry, how do I achieve that when loading models directly from HF? Do I create a model file with? ``` FROM hf.co/mradermacher/MiniMax-M2.1-i1-GGUF:Q4_K_M ``` The model exists now in the registry and I can use that. But for future models..
GiteaMirror commented 2026-05-04 23:57:39 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 19, 2026):

Create a Modelfile with FROM and TEMPLATE components. If the model is new and doesn't have an existing template that can be re-used, translate the Jinja template (usually found in the original model as chat_template.json, chat_template.jinja or the chat_template field in tokenizer_config.json) to a Go template.

<!-- gh-comment-id:3765980051 --> @rick-github commented on GitHub (Jan 19, 2026): Create a Modelfile with [`FROM`](https://github.com/ollama/ollama/blob/main/docs/modelfile.mdx#from-required) and [`TEMPLATE`](https://github.com/ollama/ollama/blob/main/docs/modelfile.mdx#template) components. If the model is new and doesn't have an existing template that can be re-used, translate the Jinja template (usually found in the original model as `chat_template.json`, `chat_template.jinja` or the `chat_template` field in `tokenizer_config.json`) to a [Go template](https://pkg.go.dev/text/template).
GiteaMirror commented 2026-05-04 23:57:40 -05:00
Author
Owner
Copy Link

@Raboo commented on GitHub (Jan 19, 2026):

hmm, https://ollama.com/frob/minimax-m2.1:latest/blobs/6dc1fe3f1a44 differs some what from https://huggingface.co/MiniMaxAI/MiniMax-M2.1/blob/main/chat_template.jinja.

<!-- gh-comment-id:3769754981 --> @Raboo commented on GitHub (Jan 19, 2026): hmm, https://ollama.com/frob/minimax-m2.1:latest/blobs/6dc1fe3f1a44 differs some what from https://huggingface.co/MiniMaxAI/MiniMax-M2.1/blob/main/chat_template.jinja.
GiteaMirror commented 2026-05-04 23:57:41 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Jan 19, 2026):

One is a Go template, the other is a Jinja template. The tool handling capability of a Go template is not flexible enough for some models, which is why some of the newer models have a parser rather than a template. Since there is no parser for minimax-m2 yet, the template is an approximate solution.

<!-- gh-comment-id:3769765120 --> @rick-github commented on GitHub (Jan 19, 2026): One is a Go template, the other is a Jinja template. The tool handling capability of a Go template is not flexible enough for some models, which is why some of the newer models have a parser rather than a template. Since there is no parser for minimax-m2 yet, the template is an approximate solution.
GiteaMirror commented 2026-05-04 23:57:42 -05:00
Author
Owner
Copy Link

@Raboo commented on GitHub (Jan 19, 2026):

ok, thanks for the explanation.

<!-- gh-comment-id:3769839829 --> @Raboo commented on GitHub (Jan 19, 2026): ok, thanks for the explanation.
GiteaMirror commented 2026-05-04 23:57:43 -05:00
Author
Owner
Copy Link

@yoisel commented on GitHub (May 4, 2026):

@Raboo I expect ollama to fetch all the data it needs to run a HF model, if HF models are claimed to be supported.
This is real bug, why was it closed?

<!-- gh-comment-id:4367750727 --> @yoisel commented on GitHub (May 4, 2026): @Raboo I expect ollama to fetch all the data it needs to run a HF model, if HF models are claimed to be supported. This is real bug, why was it closed?
GiteaMirror commented 2026-05-04 23:57:44 -05:00
Author
Owner
Copy Link

@Raboo commented on GitHub (May 4, 2026):

@yoisel I agree that the ollama way is a bit complicated. To be honest a lot of ollama choices doesn't make sense now that I know more about the LLM ecosystem.
I switched to llama.cpp and are super happy with it. Other than having better performance and better LLM support, here is a good write up on why llama.cpp is preferred for me and a lot of people.
If you want a nice one-click solutions with UI I can recommend Unsloth Studio(open source) or LM Studio(closed source). Both use llama.cpp as their inference backend like ollama did before they switched to ggml library roughly a year ago.

<!-- gh-comment-id:4370595511 --> @Raboo commented on GitHub (May 4, 2026): @yoisel I agree that the ollama way is a bit complicated. To be honest a lot of ollama choices doesn't make sense now that I know more about the LLM ecosystem. I switched to llama.cpp and are super happy with it. Other than having better performance and better LLM support, here is a good write up on why llama.cpp is preferred for me and a lot of people. If you want a nice one-click solutions with UI I can recommend Unsloth Studio(open source) or LM Studio(closed source). Both use llama.cpp as their inference backend like ollama did before they switched to ggml library roughly a year ago.
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#71083
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?