Instructions to use my0919175/Sovythos-66M-instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- llama-cpp-python
How to use my0919175/Sovythos-66M-instruct with llama-cpp-python:
# !pip install llama-cpp-python from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="my0919175/Sovythos-66M-instruct", filename="sovythos_elite.gguf", )
output = llm( "Once upon a time,", max_tokens=512, echo=True ) print(output)
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use my0919175/Sovythos-66M-instruct 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 my0919175/Sovythos-66M-instruct # Run inference directly in the terminal: llama cli -hf my0919175/Sovythos-66M-instruct
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf my0919175/Sovythos-66M-instruct # Run inference directly in the terminal: llama cli -hf my0919175/Sovythos-66M-instruct
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 my0919175/Sovythos-66M-instruct # Run inference directly in the terminal: ./llama-cli -hf my0919175/Sovythos-66M-instruct
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 my0919175/Sovythos-66M-instruct # Run inference directly in the terminal: ./build/bin/llama-cli -hf my0919175/Sovythos-66M-instruct
Use Docker
docker model run hf.co/my0919175/Sovythos-66M-instruct
- LM Studio
- Jan
- vLLM
How to use my0919175/Sovythos-66M-instruct with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "my0919175/Sovythos-66M-instruct" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "my0919175/Sovythos-66M-instruct", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/my0919175/Sovythos-66M-instruct
- Ollama
How to use my0919175/Sovythos-66M-instruct with Ollama:
ollama run hf.co/my0919175/Sovythos-66M-instruct
- Unsloth Studio
How to use my0919175/Sovythos-66M-instruct 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 my0919175/Sovythos-66M-instruct 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 my0919175/Sovythos-66M-instruct to start chatting
Using HuggingFace Spaces for Unsloth
# No setup required # Open https://huggingface.co/spaces/unsloth/studio in your browser # Search for my0919175/Sovythos-66M-instruct to start chatting
- Atomic Chat new
- Docker Model Runner
How to use my0919175/Sovythos-66M-instruct with Docker Model Runner:
docker model run hf.co/my0919175/Sovythos-66M-instruct
- Lemonade
How to use my0919175/Sovythos-66M-instruct with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull my0919175/Sovythos-66M-instruct
Run and chat with the model
lemonade run user.Sovythos-66M-instruct-{{QUANT_TAG}}List all available models
lemonade list
- π¦ Overview
- β οΈ Experimental Release (v1)
- π― Proof of Concept
- π Experience the Sovereign Platform
- π Model Specifications
- π Languages
- π» Intended Uses
- β οΈ Known Limitations
- ποΈ Built From Scratch
- πͺπ¬ Made in Egypt
- π£οΈ Roadmap
- π Future Vision
- π Citation
- π¨βπ» Creator
- π License
π¦ SOVYTHOS-66M-Instruct (v1)
An experimental instruction-tuned language model built completely from scratch under the Sovereign AI Project.
π¦ Overview
SOVYTHOS-66M-Instruct is the very first experimental instruction-tuned release of the SOVYTHOS language model family.
Unlike many open-source projects that begin from existing checkpoints, SOVYTHOS was built entirely from randomly initialized weights using a custom tokenizer and a custom training pipeline.
This release represents the first Proof of Concept (PoC) demonstrating that the SOVYTHOS architecture can learn instruction-following capabilities from scratch.
πͺπ¬ Proudly designed and trained in Egypt.
β οΈ Experimental Release (v1)
This repository contains an early experimental version.
It is NOT a stable release.
You should expect:
- Hallucinations
- Broken generations
- Incomplete answers
- Incorrect reasoning
- Context drift
- Repetition
- Mixed-language outputs
- Inconsistent coding ability
This behavior is completely expected.
The goal of this release is to demonstrate the project's progress rather than provide a production-ready assistant.
π― Proof of Concept
SOVYTHOS-66M-Instruct exists primarily to prove that:
- A language model can be built completely from scratch.
- Instruction tuning can be performed without relying on pretrained foundation models.
- The Sovereign training pipeline works end-to-end.
- Future larger models can be built upon the same foundation.
This is Version 1 of a much larger roadmap.
π Experience the Sovereign Platform
The public model is only one part of the Sovereign ecosystem.
The complete Sovereign Platform combines multiple AI systems and provides a modern AI experience beyond this standalone model.
π Model Specifications
| Property | Value |
|---|---|
| Model | SOVYTHOS-66M-Instruct |
| Parameters | ~66.7 Million |
| Architecture | Decoder-only Transformer |
| Hidden Size | 512 |
| Layers | 12 |
| Attention Heads | 16 |
| Framework | PyTorch |
| Tokenizer | Custom |
| Training | From Scratch |
| Fine-tuning | Instruction Tuning |
| Status | Experimental |
| Version | v1 |
π Languages
Current multilingual support includes:
- πͺπ¬ Arabic
- πͺπ¬ Egyptian Arabic
- πΊπΈ English
- Programming Languages
- Technical Documentation
- Mathematical Content
- General Text
Support quality is still under active development.
π» Intended Uses
This model is intended for:
- Research
- Experiments
- Community Feedback
- Educational Purposes
- Benchmarking
- Continued Training
- Fine-tuning
It is not recommended for production environments.
β οΈ Known Limitations
Because this is an early instruction model, it currently struggles with:
- Long conversations
- Mathematical reasoning
- Accurate factual recall
- Multi-step reasoning
- Reliable code generation
- Stable instruction following
- Context consistency
Future releases aim to improve all of these areas.
ποΈ Built From Scratch
SOVYTHOS was developed using:
- Custom Tokenizer
- Custom Training Pipeline
- PyTorch
- CUDA
- Mixed Precision Training
- Automatic Checkpoint Selection
No pretrained language model weights were used.
Everything starts from randomly initialized parameters.
πͺπ¬ Made in Egypt
SOVYTHOS is an independent Egyptian AI research project developed under the Sovereign AI Project.
Its long-term goal is to contribute to open-source AI with a strong focus on:
- Arabic Language
- Egyptian Arabic
- Programming
- Open Research
- Accessible AI Technology
π£οΈ Roadmap
β SOVYTHOS-66M-Base
π‘ SOVYTHOS-66M-Instruct (Experimental v1)
β¬ Stable Instruct Release
β¬ Better Arabic Understanding
β¬ Better English Understanding
β¬ Better Coding
β¬ Better Reasoning
β¬ Longer Context
β¬ 150M Model
β¬ Larger Foundation Models
β¬ Sovereign Native AI Models
π Future Vision
This repository is only the beginning.
Future generations of SOVYTHOS will focus on:
- Better reasoning
- Cleaner instruction following
- Improved multilingual understanding
- Stronger programming abilities
- Larger parameter models
- Better efficiency
- More reliable outputs
The long-term objective is to build an independent family of open-source language models developed under the Sovereign AI Project.
π Citation
@misc{sovythos66minstruct,
title={SOVYTHOS-66M-Instruct},
author={Mahmoud Yasser},
year={2026},
publisher={Hugging Face},
}
π¨βπ» Creator
Mahmoud Yasser
Founder of the Sovereign AI Project
π License
Apache License 2.0
π¦ Every model starts somewhere.
SOVYTHOS-66M-Instruct v1 is an experimental milestone and a proof that building a language model completely from scratch is possible.
It is not the destinationβit is the first step toward a larger family of Sovereign AI models.
Thank you for being part of the journey.
β If you find this project interesting, consider starring the repository and following future releases.
- Downloads last month
- -