Instructions to use sodan/dan-omni-smolm2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use sodan/dan-omni-smolm2 with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf sodan/dan-omni-smolm2
# Run inference directly in the terminal:
llama cli -hf sodan/dan-omni-smolm2

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf sodan/dan-omni-smolm2
# Run inference directly in the terminal:
llama cli -hf sodan/dan-omni-smolm2

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf sodan/dan-omni-smolm2
# Run inference directly in the terminal:
./llama-cli -hf sodan/dan-omni-smolm2

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf sodan/dan-omni-smolm2
# Run inference directly in the terminal:
./build/bin/llama-cli -hf sodan/dan-omni-smolm2

Use Docker

docker model run hf.co/sodan/dan-omni-smolm2

LM Studio
Jan

vLLM

How to use sodan/dan-omni-smolm2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "sodan/dan-omni-smolm2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "sodan/dan-omni-smolm2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/sodan/dan-omni-smolm2

Ollama
How to use sodan/dan-omni-smolm2 with Ollama:
```
ollama run hf.co/sodan/dan-omni-smolm2
```
Unsloth Desktop
Docker Model Runner
How to use sodan/dan-omni-smolm2 with Docker Model Runner:
```
docker model run hf.co/sodan/dan-omni-smolm2
```

Lemonade

How to use sodan/dan-omni-smolm2 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull sodan/dan-omni-smolm2

Run and chat with the model

lemonade run user.dan-omni-smolm2-{{QUANT_TAG}}

List all available models

lemonade list

Atomic Chat

dan-omni-smolm2

Ultra-lightweight model based on SmolLM2. Designed for extreme edge deployment — runs comfortably on any device with 512 MB RAM. Fastest in the Dan Omni family at 62 tok/s.

Model Description

Property	Value
Base Model	HuggingFace SmolLM2
Fine-tuning	Full fine-tune on mobile instruction data
Context Length	4096 tokens
Parameters	~1.7B
File Size	259 MB
RAM Usage	~400 MB at inference

System Prompt

You are dan, a helpful AI assistant for mobile devices. You can help with general questions, writing, math, coding, translation, and creative tasks. Keep answers concise and natural. Don't explain your architecture unless asked.

Benchmarks

Tested on Intel i9-9880H @ 2.30GHz, 16GB RAM, Ollama runtime.

Category	Avg tok/s	Prompt tok/s	Tokens	Time
Reasoning	60.0	545.1	26	0.4s
Coding	62.3	545.1	43	0.7s
Creative Writing	61.2	545.1	79	1.3s
Instruction Following	62.4	545.1	56	0.9s
Math	63.6	545.1	26	0.4s
General Knowledge	63.6	545.1	26	0.4s
Average	62.2	545.1	43	0.7s

Comparison vs Competitors

Model	Size	Speed	Quality	RAM
dan-omni-smolm2	259 MB	62.2 tok/s	Medium	~400 MB
SmolLM2-135M (base)	140 MB	~80 tok/s	Low	~250 MB
LFM2-1.2B	~0.7 GB	~45 tok/s	Medium-High	~1 GB
Gemma 4 E2B	~1.4 GB	~35 tok/s	High	~2 GB
dan-omni-3b	2.0 GB	11.3 tok/s	High	~2.5 GB

Why dan-omni-smolm2? 5.5x faster than dan-omni-3b, 12x smaller. Fine-tuned for instruction following — not just a base model. Runs on Raspberry Pi, old phones, IoT devices.

Speed Comparison

Model	Parameters	Size	Speed	Quality
dan-omni-3b	3B	2.0 GB	11.3 tok/s	High
dan-omni-3b-q3s	3B	1.5 GB	12.8 tok/s	Med-High
dan-omni-smolm2	~1.7B	259 MB	62.2 tok/s	Medium

Dan Omni Model Family

Model	Size	Speed	Use Case
dan-omni-3b	3.5 GB	11.3 tok/s	Full multimodal (text + vision)
dan-omni-3b-mobile	1.2 GB	9.8 tok/s	Compressed for mobile, 2K context
dan-omni-3b-q3s	1.5 GB	12.8 tok/s	Aggressive quantization
dan-omni-smolm2	259 MB	62.2 tok/s	Ultralight, fastest
dan-omni-smolm2-v2	259 MB	62.1 tok/s	Improved quality variant

Usage

ollama pull sodan/dan-omni-smolm2
ollama run sodan/dan-omni-smolm2

./llama-cli -m dan-omni-smolm2.gguf -p "Hello" --ctx-size 4096

Intended Use

Raspberry Pi / Arduino-class devices
Old phones with limited RAM
Real-time chat requiring <100ms response time
Classification, simple Q&A, translation
Prototyping and experimentation

Limitations

Noticeably weaker reasoning than 3B models
May struggle with multi-step math
Shorter, simpler responses by nature
Best paired with a larger model in a cascade

License

Apache 2.0

Downloads last month: -

GGUF

Model size

0.4B params

Architecture

llama

Hardware compatibility

We're not able to determine the quantization variants.

View all variants