github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #12126] [Model Request] NVIDIA-Nemotron-Nano-9B-v2 #54570

New Issue
Closed
opened 2026-04-29 06:23:15 -05:00 by GiteaMirror · 11 comments
GiteaMirror commented 2026-04-29 06:23:15 -05:00
Owner
Copy Link

Originally created by @BoguesUser on GitHub (Aug 30, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/12126

Nemotron support has been added upstream to llama.cpp in b6316

Any plans to also add it to ollama?

Originally created by @BoguesUser on GitHub (Aug 30, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/12126 Nemotron support has been added upstream to llama.cpp [in b6316](https://github.com/ggml-org/llama.cpp/releases/tag/b6315) Any plans to also add it to ollama?
GiteaMirror added the model label 2026-04-29 06:23:15 -05:00
GiteaMirror commented 2026-04-29 06:23:18 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Aug 30, 2025):

It will be added when the next vendor sync happens.

<!-- gh-comment-id:3239413865 --> @rick-github commented on GitHub (Aug 30, 2025): It will be added when the next vendor sync happens.
GiteaMirror commented 2026-04-29 06:23:18 -05:00
Author
Owner
Copy Link

@NEWbie0709 commented on GitHub (Sep 2, 2025):

Can I know when the next vendor sync will be?

<!-- gh-comment-id:3243750306 --> @NEWbie0709 commented on GitHub (Sep 2, 2025): Can I know when the next vendor sync will be?
GiteaMirror commented 2026-04-29 06:23:19 -05:00
Author
Owner
Copy Link

@donhuvy commented on GitHub (Sep 3, 2025):

I experiment NVIDIA Nemotron 9B v2 today, it is great reasoning model with only 9B parameters. Let's integration it to Olama.

<!-- gh-comment-id:3247519421 --> @donhuvy commented on GitHub (Sep 3, 2025): I experiment NVIDIA Nemotron 9B v2 today, it is great reasoning model with only 9B parameters. Let's integration it to Olama.
GiteaMirror commented 2026-04-29 06:23:20 -05:00
Author
Owner
Copy Link

@NEWbie0709 commented on GitHub (Sep 5, 2025):

I think Ollama was updated today, but it still doesn’t support the Nemotron model.

<!-- gh-comment-id:3257027515 --> @NEWbie0709 commented on GitHub (Sep 5, 2025): I think Ollama was updated today, but it still doesn’t support the Nemotron model.
GiteaMirror commented 2026-04-29 06:23:21 -05:00
Author
Owner
Copy Link

@ithax-wb commented on GitHub (Sep 8, 2025):

It’s a hybrid model. Can you also support the think and no_think option in the API.

P.S There is also a 12B variant of this model

<!-- gh-comment-id:3267015000 --> @ithax-wb commented on GitHub (Sep 8, 2025): It’s a hybrid model. Can you also support the think and no_think option in the API. P.S There is also a 12B variant of this model
GiteaMirror commented 2026-04-29 06:23:21 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Sep 14, 2025):

Vendor sync in progress in https://github.com/ollama/ollama/pull/12245.

<!-- gh-comment-id:3289950757 --> @rick-github commented on GitHub (Sep 14, 2025): Vendor sync in progress in https://github.com/ollama/ollama/pull/12245.
GiteaMirror commented 2026-04-29 06:23:22 -05:00
Author
Owner
Copy Link

@Chovus13 commented on GitHub (Oct 7, 2025):

Please friends, add this model to Ollama. Thanks in advance

<!-- gh-comment-id:3374755323 --> @Chovus13 commented on GitHub (Oct 7, 2025): Please friends, add this model to Ollama. Thanks in advance
GiteaMirror commented 2026-04-29 06:23:23 -05:00
Author
Owner
Copy Link

@jozzo402 commented on GitHub (Oct 14, 2025):

Vendor sync in progress in #12245.

The linked pull request has been merged, is this model supported now?

<!-- gh-comment-id:3401271091 --> @jozzo402 commented on GitHub (Oct 14, 2025): > Vendor sync in progress in [#12245](https://github.com/ollama/ollama/pull/12245). The linked pull request has been merged, is this model supported now?
GiteaMirror commented 2026-04-29 06:23:24 -05:00
Author
Owner
Copy Link

@harbachonak commented on GitHub (Oct 17, 2025):

Hi! @rick-github Are there any updates on this? Can't find the model yet

<!-- gh-comment-id:3414869941 --> @harbachonak commented on GitHub (Oct 17, 2025): Hi! @rick-github Are there any updates on this? Can't find the model yet
GiteaMirror commented 2026-04-29 06:23:25 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Oct 17, 2025):

The model can be pulled from hf.co/bartowski/nvidia_NVIDIA-Nemotron-Nano-9B-v2-GGUF. However, the template for that model is not great, I'm testing a rewrite and will post it here shortly.

<!-- gh-comment-id:3414944030 --> @rick-github commented on GitHub (Oct 17, 2025): The model can be pulled from hf.co/bartowski/nvidia_NVIDIA-Nemotron-Nano-9B-v2-GGUF. However, the template for that model is not great, I'm testing a rewrite and will post it here shortly.
GiteaMirror commented 2026-04-29 06:23:26 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Oct 17, 2025):

FROM hf.co/bartowski/nvidia_NVIDIA-Nemotron-Nano-9B-v2-GGUF:Q6_K_L
TEMPLATE """<SPECIAL_10>System
{{- if or .System .Tools }}
{{- if .System }}
{{ .System }}
{{ end }}
{{- if .Tools }}
You can use the following tools to assist the user if required:
<AVAILABLE_TOOLS>[
{{- range $i, $tool := $.Tools }}
{{- $last := eq (len (slice $.Tools $i)) 1 -}}
{"description": "{{ .Function.Description }}", "name": "{{ .Function.Name }}", "parameters": {"properties": {{ .Function.Parameters.Properties |json }}, "required": {{ .Function.Parameters.Required |json }} } }
{{- if not $last }}{{ `, ` }}
{{- end }}
{{- end -}}
]</AVAILABLE_TOOLS>

If you decide to call any tool(s), use the following format:
<TOOLCALL>[{"name": "tool_name1", "arguments": "tool_args1"}, {"name": "tool_name2", "arguments": "tool_args2"}]</TOOLCALL>

The user will execute tool-calls and return responses from tool(s) in this format:
<TOOL_RESPONSE>[{"tool_response1"}, {"tool_response2"}]</TOOL_RESPONSE>

Based on the tool responses, you can call additional tools if needed, correct tool calls if any errors are found, or just respond to the user.
{{ end }}
{{- end }}

{{- range $i, $_ := .Messages }}
{{- $last := eq (len (slice $.Messages $i)) 1}}
{{- if eq .Role "user" }}<SPECIAL_11>User
{{ .Content }}
{{ else if eq .Role "tool" }}<SPECIAL_11>User
<TOOL_RESPONSE>[
{{- .Content -}}
]</TOOL_RESPONSE>
{{ else if eq .Role "assistant" }}<SPECIAL_11>Assistant
{{- if and $.IsThinkSet (and $last and (.Thinking false)) -}}
<think>
{{ .Thinking }}
</think>
{{ end }}
{{- if .Content }}
{{ .Content }}
{{- end }}
{{- if .ToolCalls }}
<TOOLCALL>[
{{- range $i, $_ := .ToolCalls }}{"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}{{ end -}}
]</TOOLCALL>
{{- end }}
<SPECIAL_12>
{{ end }}
{{- if and $last (ne .Role "assistant") }}<SPECIAL_11>Assistant
<think>
{{- if and $.IsThinkSet (not $.Think) -}}
</think>
{{- end }}
{{- end }}
{{- end }}
"""
PARAMETER stop <SPECIAL_10>
PARAMETER stop <SPECIAL_11>
PARAMETER stop <SPECIAL_12>
PARAMETER stop <think>

Save this as Modelfile. Then run:

ollama pull hf.co/bartowski/nvidia_NVIDIA-Nemotron-Nano-9B-v2-GGUF:Q6_K_L
ollama create nemotron-nano:9b-v2-q6_K_L -f Modelfile

Feel free to change Q6_K_L/q6_K_L to a different quantization.

The modelfile adds thinking and tool support and in my quick tests is seems to work fine, but YMMV.

$ ollama run nemotron-nano:9b-v2-q6_K_L hello
Thinking...
Okay, the user just said "hello". That's a greeting. I should respond politely and offer help. Let me make sure my response is friendly and open-ended so they feel comfortable asking questions or sharing what's on their mind.
...done thinking.

Hello! How can I assist you today? 😊


$ ollama run nemotron-nano:9b-v2-q6_K_L hello --think=false
Hello! How can I assist you today? 😊

$ ./ollama-run.py nemotron-nano:9b-v2-q6_K_L what time is it --tool get_datetime
Thinking...
Okay, the user asked "what time is it". I need to provide the current time. The available tool is get_datetime, which returns the current date and time in a specific format. Since the user didn't specify a timezone, I should use the local time as per the tool's description. The parameters for get_datetime include an optional timezone_name, but since it's not required and the user didn't mention any specific timezone, I can leave it blank or assume local time. So I'll call get_datetime without specifying the timezone to get the current local time.

...done thinking

calling get_datetime({'timezone_name': ''})
Thinking...
Okay, the user asked "what time is it". I need to provide the current time. The tool get_datetime was called without a timezone, so it returned local time. The response from the tool shows the time as 18:29. I should format this in a friendly way. Let me check the date and time details again. The fulldate is Friday, October 17, 2025 at 18:29. Since the user didn't specify a timezone, using local time is correct. I'll just state the current time clearly.

...done thinking
The current time is **18:29** (6:29 PM) on Friday, October 17, 2025. Let me know if you need help with anything else!
<!-- gh-comment-id:3416288252 --> @rick-github commented on GitHub (Oct 17, 2025): ```dockerfile FROM hf.co/bartowski/nvidia_NVIDIA-Nemotron-Nano-9B-v2-GGUF:Q6_K_L TEMPLATE """<SPECIAL_10>System {{- if or .System .Tools }} {{- if .System }} {{ .System }} {{ end }} {{- if .Tools }} You can use the following tools to assist the user if required: <AVAILABLE_TOOLS>[ {{- range $i, $tool := $.Tools }} {{- $last := eq (len (slice $.Tools $i)) 1 -}} {"description": "{{ .Function.Description }}", "name": "{{ .Function.Name }}", "parameters": {"properties": {{ .Function.Parameters.Properties |json }}, "required": {{ .Function.Parameters.Required |json }} } } {{- if not $last }}{{ `, ` }} {{- end }} {{- end -}} ]</AVAILABLE_TOOLS> If you decide to call any tool(s), use the following format: <TOOLCALL>[{"name": "tool_name1", "arguments": "tool_args1"}, {"name": "tool_name2", "arguments": "tool_args2"}]</TOOLCALL> The user will execute tool-calls and return responses from tool(s) in this format: <TOOL_RESPONSE>[{"tool_response1"}, {"tool_response2"}]</TOOL_RESPONSE> Based on the tool responses, you can call additional tools if needed, correct tool calls if any errors are found, or just respond to the user. {{ end }} {{- end }} {{- range $i, $_ := .Messages }} {{- $last := eq (len (slice $.Messages $i)) 1}} {{- if eq .Role "user" }}<SPECIAL_11>User {{ .Content }} {{ else if eq .Role "tool" }}<SPECIAL_11>User <TOOL_RESPONSE>[ {{- .Content -}} ]</TOOL_RESPONSE> {{ else if eq .Role "assistant" }}<SPECIAL_11>Assistant {{- if and $.IsThinkSet (and $last and (.Thinking false)) -}} <think> {{ .Thinking }} </think> {{ end }} {{- if .Content }} {{ .Content }} {{- end }} {{- if .ToolCalls }} <TOOLCALL>[ {{- range $i, $_ := .ToolCalls }}{"name": "{{ .Function.Name }}", "arguments": {{ .Function.Arguments }}{{ end -}} ]</TOOLCALL> {{- end }} <SPECIAL_12> {{ end }} {{- if and $last (ne .Role "assistant") }}<SPECIAL_11>Assistant <think> {{- if and $.IsThinkSet (not $.Think) -}} </think> {{- end }} {{- end }} {{- end }} """ PARAMETER stop <SPECIAL_10> PARAMETER stop <SPECIAL_11> PARAMETER stop <SPECIAL_12> PARAMETER stop <think> ``` Save this as `Modelfile`. Then run: ``` ollama pull hf.co/bartowski/nvidia_NVIDIA-Nemotron-Nano-9B-v2-GGUF:Q6_K_L ollama create nemotron-nano:9b-v2-q6_K_L -f Modelfile ``` Feel free to change Q6_K_L/q6_K_L to a different quantization. The modelfile adds thinking and tool support and in my quick tests is seems to work fine, but YMMV. ```console $ ollama run nemotron-nano:9b-v2-q6_K_L hello Thinking... Okay, the user just said "hello". That's a greeting. I should respond politely and offer help. Let me make sure my response is friendly and open-ended so they feel comfortable asking questions or sharing what's on their mind. ...done thinking. Hello! How can I assist you today? 😊 $ ollama run nemotron-nano:9b-v2-q6_K_L hello --think=false Hello! How can I assist you today? 😊 $ ./ollama-run.py nemotron-nano:9b-v2-q6_K_L what time is it --tool get_datetime Thinking... Okay, the user asked "what time is it". I need to provide the current time. The available tool is get_datetime, which returns the current date and time in a specific format. Since the user didn't specify a timezone, I should use the local time as per the tool's description. The parameters for get_datetime include an optional timezone_name, but since it's not required and the user didn't mention any specific timezone, I can leave it blank or assume local time. So I'll call get_datetime without specifying the timezone to get the current local time. ...done thinking calling get_datetime({'timezone_name': ''}) Thinking... Okay, the user asked "what time is it". I need to provide the current time. The tool get_datetime was called without a timezone, so it returned local time. The response from the tool shows the time as 18:29. I should format this in a friendly way. Let me check the date and time details again. The fulldate is Friday, October 17, 2025 at 18:29. Since the user didn't specify a timezone, using local time is correct. I'll just state the current time clearly. ...done thinking The current time is **18:29** (6:29 PM) on Friday, October 17, 2025. Let me know if you need help with anything else! ```
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 model
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#54570
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?