Instructions to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="EXAiStudio/EX-Ai1-1.72B-Qwen3")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("EXAiStudio/EX-Ai1-1.72B-Qwen3", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 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 EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama cli -hf EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

Use Docker

docker model run hf.co/EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

Llama new
LM Studio
Jan

vLLM

How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "EXAiStudio/EX-Ai1-1.72B-Qwen3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "EXAiStudio/EX-Ai1-1.72B-Qwen3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

SGLang

How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "EXAiStudio/EX-Ai1-1.72B-Qwen3" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "EXAiStudio/EX-Ai1-1.72B-Qwen3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "EXAiStudio/EX-Ai1-1.72B-Qwen3" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "EXAiStudio/EX-Ai1-1.72B-Qwen3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with Ollama:
```
ollama run hf.co/EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M
```
Unsloth Desktop

How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with Docker Model Runner:
```
docker model run hf.co/EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M
```

Lemonade

How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

Run and chat with the model

lemonade run user.EX-Ai1-1.72B-Qwen3-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

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 EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use EXAiStudio/EX-Ai1-1.72B-Qwen3 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

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 "EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M" \
  --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"

EX-Ai1-1.72B-Qwen3

EX-Ai1-1.72B-Qwen3 is a LoRA fine-tuned version of Qwen3-1.7B, developed by EXAiStudio. It is optimized for Persian language understanding, mathematical reasoning, and instruction-following, while retaining multilingual capabilities in English, Chinese, and Russian.

The model is released as GGUF (Q4_K_M) for efficient local inference on consumer hardware.

📌 Model Details

Property	Value
Developed by	EXAiStudio
Model type	Causal Language Model (Decoder-only Transformer)
Base model	Qwen/Qwen3-1.7B
Parameters	1.72B
Architecture	Qwen3 (24 layers, hidden size 2048, 16 attention heads)
Context length	32,768 tokens (same as base model, natively supported)
Fine-tuning method	LoRA
Languages	Persian (fa), English (en), Chinese (zh), Russian (ru)
License	Apache-2.0
Version	V1.0
Release date	October 16, 2026
Format	GGUF (Q4_K_M only)

Description

EX-Ai1-1.72B-Qwen3 is a lightweight, efficient language model derived from Qwen3-1.7B. It was fine-tuned with a LoRA adapter using a small, curated, proprietary dataset of ~2,000 samples developed internally at EXAiStudio, focused on improving:

Persian (Farsi) natural language understanding and generation
Mathematical reasoning and step-by-step problem solving
Instruction-following and structured responses

The resulting model runs comfortably on mid-range GPUs (e.g., T4, RTX 3060, or Apple Silicon) with the Q4_K_M quantization.

🎯 Intended Use

✅ Direct Use

Persian and multilingual conversational assistants
Math problem solving and step-by-step explanations
Instruction-following tasks (Q&A, summarization, rewriting)
Local/offline deployment via Ollama, llama.cpp, or LM Studio

❌ Out-of-Scope Use

Advanced mathematics and complex calculations beyond standard reasoning
Sensitive advice (medical, legal, financial, psychological)
Full reliance on model output as factual truth
High-stakes automated decision-making
Safety-critical systems

🏋️ Training Details

Training Data

Source: Proprietary dataset developed by EXAiStudio
Size: ~2,000 samples
Focus areas: Persian language, mathematics, instruction-following
Max sequence length: same as base model (32,768 tokens)

Training Procedure

Method: LoRA fine-tuning
Hardware: 1× NVIDIA Tesla T4 (16 GB VRAM)
Training time: ~1 hour
Base model: Qwen3-1.7B (full precision)

Detailed hyperparameters (LoRA rank, alpha, learning rate, epochs, optimizer) are not disclosed at this time.

Quantization

Only a single quantized format is provided:

Q4_K_M (GGUF) — for llama.cpp, Ollama, LM Studio, and compatible runtimes.

💻 How to Use

With Ollama

ollama run hf.co/EXAiStudio/EX-Ai1-1.72B-Qwen3:Q4_K_M

Downloads last month: 267

GGUF

Model size

2B params

Architecture

qwen3

Hardware compatibility

4-bit

Model tree for EXAiStudio/EX-Ai1-1.72B-Qwen3

Base model

Qwen/Qwen3-1.7B-Base

Finetuned

Qwen/Qwen3-1.7B

Finetuned

(1356)

this model

Collection including EXAiStudio/EX-Ai1-1.72B-Qwen3

EX Ai Studio

Collection

1 item • Updated 8 days ago