Instructions to use MaxKio/Mio-1.0-Pro with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MaxKio/Mio-1.0-Pro with Transformers:

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

pipe = pipeline("text-generation", model="MaxKio/Mio-1.0-Pro")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("MaxKio/Mio-1.0-Pro")
model = AutoModelForCausalLM.from_pretrained("MaxKio/Mio-1.0-Pro")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use MaxKio/Mio-1.0-Pro with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "MaxKio/Mio-1.0-Pro"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "MaxKio/Mio-1.0-Pro",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/MaxKio/Mio-1.0-Pro

SGLang

How to use MaxKio/Mio-1.0-Pro 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 "MaxKio/Mio-1.0-Pro" \
    --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": "MaxKio/Mio-1.0-Pro",
		"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 "MaxKio/Mio-1.0-Pro" \
        --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": "MaxKio/Mio-1.0-Pro",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use MaxKio/Mio-1.0-Pro with Docker Model Runner:
```
docker model run hf.co/MaxKio/Mio-1.0-Pro
```

Mio-1.0-Pro / README.md

MaxKio

Upload README.md with huggingface_hub

e4c040a verified 3 days ago

preview code

Raw

History Blame Contribute Delete

2.76 kB

	---
	license: apache-2.0
	language:
	- en
	- ar
	- zh
	- es
	- fr
	- de
	- ru
	- ja
	- ko
	- pt
	library_name: transformers
	tags:
	- qwen2
	- chat
	- instruct
	- 128k-context
	- lightweight
	- multilingual
	base_model: Qwen/Qwen2.5-0.5B-Instruct
	pipeline_tag: text-generation
	---

	# Mio 1.0 Pro

	<div align="center">

	Advanced Lightweight AI Assistant \| 494M Parameters \| 128K Context Window

	</div>

	## Overview

	Mio 1.0 Pro is a lightweight yet powerful AI assistant model based on Qwen2.5-0.5B-Instruct, enhanced with extended context support and optimized for responsive, high-quality conversations. Designed to run efficiently on resource-constrained environments including CPU-only deployments.

	## Key Features

	- 494M Parameters - Ultra-lightweight, runs on CPU with ~1.4GB RAM
	- 128K Context Window - Extended context via RoPE scaling (rope_theta: 4,000,000)
	- Multilingual - Supports 20+ languages including Arabic, English, Chinese, and more
	- Code Generation - Enhanced in-context learning for programming tasks
	- CPU Optimized - Runs efficiently without GPU acceleration

	## Modifications from Base Model

	\| Feature \| Base (Qwen2.5-0.5B) \| Mio 1.0 Pro \|
	\|---------\|---------------------\|-------------\|
	\| Context Length \| 32K \| 128K \|
	\| RoPE Theta \| 1,000,000 \| 4,000,000 \|
	\| In-Context Learning \| Default \| Enhanced code examples \|

	## Quick Start

	```python
	from transformers import AutoModelForCausalLM, AutoTokenizer

	model_name = "MaxKio/Mio-1.0-Pro"

	tokenizer = AutoTokenizer.from_pretrained(model_name)
	model = AutoModelForCausalLM.from_pretrained(
	model_name,
	dtype="auto",
	device_map="auto"
	)

	messages = [
	{"role": "system", "content": "You are Mio 1.0 Pro, an advanced AI assistant."},
	{"role": "user", "content": "Write a Python function to check if a number is prime."}
	]

	text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
	inputs = tokenizer(text, return_tensors="pt")
	outputs = model.generate(**inputs, max_new_tokens=512)
	print(tokenizer.decode(outputs[0], skip_special_tokens=True))
	```

	## Deployment

	Mio 1.0 Pro is optimized for lightweight server deployment:

	```bash
	# Minimum requirements
	# - RAM: 2GB
	# - Disk: 1GB
	# - CPU: Any modern x86/ARM processor
	# - GPU: Optional (not required)
	```

	## Supported Languages

	English, Arabic, Chinese (Simplified/Traditional), Spanish, French, German, Russian, Japanese, Korean, Portuguese, Italian, Dutch, Polish, Turkish, Vietnamese, Thai, Indonesian, Hindi, and more.

	## License

	Apache 2.0 - This model is built upon [Qwen2.5-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct) by Qwen/Alibaba, licensed under Apache 2.0.

	## Author

	MaxKio - [HuggingFace Profile](https://huggingface.co/MaxKio)