theOG-50M

theOG-50M

Writes a Python function from a plain English request. A 50M parameter code model trained from scratch on one laptop, small enough to run offline in Ollama, LM Studio or llama.cpp.

Made by DrunkkToys on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: theOG-50M v3 is a Python code model with about 50M parameters, trained from scratch by one developer, DrunkkToys. You type a short English description of the function you want, and the model replies with Python code. The download is 54 MB in GGUF format, so it runs offline on an ordinary laptop.

Who it suits: Programmers and tinkerers who want a tiny local model for short Python snippets, and anyone curious about what a small model trained on one laptop can do.

Industry | Who uses it | How they use it | Example
Software development | Python programmers | Ask for small helper functions offline | Requesting a function that reverses a string
Education | Students who learn Python | Compare the model's code with their own | Checking an alternative way to write a loop
Hobby and research | People who study small models | Run a 50M code model and inspect its limits | Testing it on a set of simple coding prompts

Skip it if you need long programs or other languages, since the card describes a small Python model.

Where to use it: It runs fully offline in Ollama, LM Studio or llama.cpp. The files download on Windows, macOS or Linux.

• Ollama
• LM Studio
• llama.cpp
• Windows, macOS or Linux: where you can fetch the files with Python

Use case | Who | What they get back
Function from a description | Developers | Python code for a small function
Offline coding help | Travellers and privacy minded users | Code without any network access
Small model study | Researchers | A compact model to inspect

How to set it up: You need Python to download the file and one of the apps above to run it. Only about 25 MB of weights at 4-bit are listed on Hugging Face, and the card says the download is 54 MB.

• Python with pip
• Ollama, LM Studio or llama.cpp
• A laptop with no special hardware

1. Install the Hugging Face hub library.
2. Download the model files.
3. Load the GGUF file in Ollama, LM Studio or llama.cpp.

pip install -U huggingface_hub
hf download DrunkkToys/theOG-50M

How to use it: Load the file in your app, then type what you want the function to do in plain English. The card shows no commands beyond the download, so use your app's normal way of running a GGUF file.

Tool | What it does
hf download | Fetches the model files to your computer
Ollama, LM Studio, llama.cpp | Load the GGUF file and run the model

Example: You ask it for a Python function that checks whether a number is prime. A model of this size is likely to return a short function, and you should read and test it before you rely on it.

For newcomers and experts: If you are new to this, start by loading the file in LM Studio and asking for one small function. Run the code it gives you and check the result.

If you know the field, the card says it was trained over 346,000 steps on a 2.49B token dataset, using PyTorch on the Metal GPU of a MacBook Air. It compares itself with CodeGen-350M-mono, which saw far more training data and has about seven times the parameters.

Plain English request
Small Python model reads it
It writes a Python function
You run and test the code

Good to know: The card is partly the author's personal account. It also holds a long list of complaints about an AI coding assistant that the author used, and this page does not assess those claims.

• License: Apache 2.0
• Cost to run: none named, and the card says it needs no special hardware
• Data it reaches: the prompts you give it, with no network access needed
• Not for: large programs, other languages or production code without review

On Hugging Face, 13 November 2026

• Task: text generation
• Library: gguf
• Parameters: 50.1M
• Weights: about 25 MB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 1,312
• Likes: 3
• License: apache-2.0
• Languages: en
• Tags: gguf, code, python, small-language-model, qwen3-arch, ollama, lmstudio, text-generation, conversational

Links

• Model: https://huggingface.co/DrunkkToys/theOG-50M
• Files: https://huggingface.co/DrunkkToys/theOG-50M/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download DrunkkToys/theOG-50M

Then follow the usage in its model card. Nothing in this kit runs by itself.