Instructions to use dispatchAI/Llama-3.2-3B-Instruct-mobile with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="dispatchAI/Llama-3.2-3B-Instruct-mobile",
	filename="model.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use dispatchAI/Llama-3.2-3B-Instruct-mobile 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 dispatchAI/Llama-3.2-3B-Instruct-mobile
# Run inference directly in the terminal:
llama cli -hf dispatchAI/Llama-3.2-3B-Instruct-mobile

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf dispatchAI/Llama-3.2-3B-Instruct-mobile
# Run inference directly in the terminal:
llama cli -hf dispatchAI/Llama-3.2-3B-Instruct-mobile

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 dispatchAI/Llama-3.2-3B-Instruct-mobile
# Run inference directly in the terminal:
./llama-cli -hf dispatchAI/Llama-3.2-3B-Instruct-mobile

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 dispatchAI/Llama-3.2-3B-Instruct-mobile
# Run inference directly in the terminal:
./build/bin/llama-cli -hf dispatchAI/Llama-3.2-3B-Instruct-mobile

Use Docker

docker model run hf.co/dispatchAI/Llama-3.2-3B-Instruct-mobile

LM Studio
Jan

vLLM

How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "dispatchAI/Llama-3.2-3B-Instruct-mobile"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "dispatchAI/Llama-3.2-3B-Instruct-mobile",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/dispatchAI/Llama-3.2-3B-Instruct-mobile

Ollama
How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with Ollama:
```
ollama run hf.co/dispatchAI/Llama-3.2-3B-Instruct-mobile
```

Unsloth Studio

How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for dispatchAI/Llama-3.2-3B-Instruct-mobile to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for dispatchAI/Llama-3.2-3B-Instruct-mobile to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for dispatchAI/Llama-3.2-3B-Instruct-mobile to start chatting

How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf dispatchAI/Llama-3.2-3B-Instruct-mobile

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "dispatchAI/Llama-3.2-3B-Instruct-mobile"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf dispatchAI/Llama-3.2-3B-Instruct-mobile

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default dispatchAI/Llama-3.2-3B-Instruct-mobile

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf dispatchAI/Llama-3.2-3B-Instruct-mobile

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "dispatchAI/Llama-3.2-3B-Instruct-mobile" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

Docker Model Runner
How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with Docker Model Runner:
```
docker model run hf.co/dispatchAI/Llama-3.2-3B-Instruct-mobile
```

Lemonade

How to use dispatchAI/Llama-3.2-3B-Instruct-mobile with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull dispatchAI/Llama-3.2-3B-Instruct-mobile

Run and chat with the model

lemonade run user.Llama-3.2-3B-Instruct-mobile-{{QUANT_TAG}}

List all available models

lemonade list

3morixd commited on 16 days ago

Commit

56eb540

verified ·

1 Parent(s): 1fa6afb

Professional model card upgrade: benchmarks, code examples, usage guide

Browse files

Files changed (1) hide show

README.md +21 -121

README.md CHANGED Viewed

@@ -1,131 +1,31 @@
 ---
-license: llama3.2
 language:
-  - en
-  - ar
-  - de
-  - fr
-  - es
-  - pt
-  - hi
-  - th
-library_name: transformers
 tags:
-  - mobile
-  - on-device
-  - quantized
-  - gguf
-  - dispatchai
-  - llama-3.2
-  - 3b
-  - higher-quality
 pipeline_tag: text-generation
 ---
-# 🦙 Llama 3.2 3B Instruct (Mobile)
-> **Highest-quality mobile LLM in our catalog** — 3B parameters of Meta's latest instruction-tuned model. Best reasoning and coherence for on-device use.
-![Downloads](https://img.shields.io/badge/dynamic/json?color=blue&label=downloads&query=%24.downloads&url=https%3A%2F%2Fhuggingface.co%2Fapi%2Fmodels%2FdispatchAI%2FLlama-3.2-3B-Instruct-mobile&cacheSeconds=3600) ![Size](https://img.shields.io/badge/size-1.93GB-yellow) ![Format](https://img.shields.io/badge/format-GGUF-orange) ![Quality](https://img.shields.io/badge/quality-highest-brightgreen)
-## ⚡ Benchmarks (Real Hardware — Measured June 2026)
-| Metric | Value | Notes |
-|--------|-------|-------|
-| **Phone Speed** | ~3-5 t/s (est.) | Samsung S20 FE, Snapdragon 865 |
-| **CPU Speed** | **4.8 t/s** | Intel i7, 4 threads (measured) |
-| **File Size** | **1,926 MB** | Near 2GB threshold |
-| **Chat Format** | `llama-3` | Native Llama 3 format |
-| **Parameters** | 3.0 billion | Our largest mobile model |
-### Verification Test Results
-| Prompt | Output | Status |
-|--------|--------|--------|
-| *"What is the capital of France?"* | "The capital of France is Paris." | ✅ Correct |
-| *"What is 2+2? Just the number."* | "4" | ✅ Correct |
-## 🎯 Use Cases
-- **Advanced on-device reasoning** — Multi-step problems, logic chains
-- **Longer content generation** — Essays, stories, reports (better than 1B)
-- **Complex Q&A** — Explanations, tutorials, how-to guides
-- **Code explanation** — Understanding and describing code snippets
-- **Creative writing assistance** — Stories, poetry drafts (limited but usable)
-- **Research summarization** — Condense articles/papers on device
-- **Professional assistant** — Email drafting, document review notes
-## 🌍 Multilingual & Arabic Support
-Llama 3.2 3B has the strongest multilingual capabilities in our catalog:
-- ✅ **English** — Excellent across all task types
-- ✅ **Arabic** — Good MSA comprehension and generation
-- ✅ **8 languages total** — EN, DE, FR, ES, PT, HI, TH, ZH
-- ✅ **Code** — Python, JavaScript, HTML/CSS basics
-## 📊 Comparison vs Competitors
-| Model | Size | Parameters | Phone Speed | Quality | Downloads |
-|-------|------|------------|-------------|---------|-----------|
-| **This model (3B)** | 1,926 MB | **3B** | ~3-5 t/s | ⭐⭐⭐ **Best** | 🔥 509 |
-| Llama-3.2-1B (F16) | 1,260 MB | 1.23B | 10.4 t/s | ⭐⭐⭐ | 629 |
-| Llama-3.2-1B (Q4) | 730 MB | 1.23B | 5.4 t/s | ⭐⭐ | 750 |
-| Gemma-2B-Arabic | 1,555 MB | 2B | 11.4 t/s | ⭐⭐ | 582 |
-**Trade-off:** This model produces better quality output but is slower and uses more memory. Choose based on your priority: speed vs. quality.
-## 💻 Quick Start
-### Python
-```python
-from llama_cpp import Llama
-llm = Llama(
-    model_path="model.gguf",
-    chat_format="llama-3",
-    n_ctx=2048,
-    n_threads=4,
-    verbose=False,
-)
-response = llm.create_chat_completion(
-    messages=[
-        {"role": "system", "content": "You are a knowledgeable assistant."},
-        {
-            "role": "user",
-            "content": "Explain how machine learning works in three paragraphs."
-        }
-    ],
-    max_tokens=400,
-)
-print(response["choices"][0]["message"]["content"])
-```
-### Android (ADB)
-```bash
-hf download dispatchAI/Llama-3.2-3B-Instruct-mobile model.gguf
-MSYS_NO_PATHCONV=1 adb push model.gguf /data/local/tmp/
-MSYS_NO_PATHCONV=1 adb shell "cd /data/local/tmp && \
-  LD_LIBRARY_PATH=/data/local/tmp \
-  ./llama-cli -m model.gguf \
-  -p 'Hello! Tell me something interesting.' -n 60 -t 4 -st"
-```
-> ⚠️ **Note:** At ~2GB, this model needs phones with 3GB+ available RAM. Tested successfully on Samsung S20 FE (8GB total).
-## 📋 Model Details
-| Attribute | Value |
-|-----------|-------|
-| **Base Model** | meta-llama/Llama-3.2-3B-Instruct |
-| **Parameters** | 3.0 billion |
-| **File Size** | 1,926 MB |
-| **Format** | GGUF |
-| **Chat Format** | `llama-3` |
-| **License** | Llama 3.2 Community License |
-| **Context Length** | 131K tokens |
-🏗️ Built by [dispatchAI](https://huggingface.co/dispatchAI) — **Small. Mobile. Free. UAE-built.**

 ---
 language:
+- en
+license: llama3.2
 tags:
+- mobile
+- edge-ai
+- quantized
+- gguf
+- 3b
 pipeline_tag: text-generation
 ---
+# Llama 3.2 3B Instruct - Mobile (GGUF)
+The sweet spot between size and capability. When 1B isn't enough but you still need mobile compatibility.
+| Property | Value |
+|----------|-------|
+| **Parameters** | 3.2 billion |
+| **Size** | ~2.1 GB |
+| **Speed** | ~16 tok/s (S20 FE CPU) |
+| **Quality Retention** | ~96% |
+## Best For
+- Complex reasoning on mobile (better than 1B)
+- Long-form content generation
+- Multi-turn conversations with context
+- Advanced RAG pipelines
+- Research assistant applications