mine:o2-small

mine:o2-small

A small 355M English chat model from MineAI Technology that follows a simple User and Assistant format. A research-grade model that stops cleanly and writes short code, with no added knowledge.

Made by MineAITechnology 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: mine:o2-small SFT is a 355M-parameter chat model from MineAI Technology, developed by Bilal Mehtab. It is a decoder-only model in the GPT-NeoX family with a 4096-token context window. It was fine-tuned in three short rounds on about 50M training tokens, and it answers in a User and Assistant format.

Who it suits: Learners and researchers who want a tiny model they can run and inspect, and developers who need a lightweight chat baseline.

Industry | Who uses it | How they use it | Example
Research | ML researchers | Study a small decoder-only chat model and its limits | A baseline for a small-model comparison
Education | Students | Learn how fine-tuning shapes behavior | A walkthrough of three SFT rounds
Software | Developers | Test a chat format on a tiny model | A prototype that stops cleanly
Hobby | Tinkerers | Run a model on modest hardware | A local chat toy

If you need accurate facts or reasoning, you can skip it, since the card describes it as a small research model.

Where to use it:

• Windows, macOS or Linux with Python, using the transformers library
• Local experiments on a laptop, given the small size
• Teaching and benchmarking setups that compare small models

Use case | Who | What they get back
Short chat answers | Individuals | Brief replies that end cleanly
Short code snippets | Learners | Small pieces of code
Safety behavior checks | Researchers | Clear refusals of plainly harmful requests

How to set it up:

• Python, to run the Hugging Face download tool
• The transformers library to load the model
• Modest memory, since the model has about 355M parameters

1. Install the Hugging Face hub tools with pip.
2. Download the model files with the hf command.
3. Load the model with transformers and follow the card for the prompt format.

pip install -U huggingface_hub
hf download MineAITechnology/mine-o2-small

How to use it: Write your message as a User turn and let the model continue as the Assistant. The card says it follows this simple chat format, stops cleanly and knows who made it.

Example: You write a User turn asking for a short Python function that adds two numbers. The model answers as the Assistant with a few lines of code and then stops, but do not rely on it for facts.

For newcomers and experts: Start with one short instruction and read the reply critically. The card is frank that this is a small research model, so treat it as a toy or a baseline.

It uses 24 Transformer blocks with a hidden size of 1024, 16 attention heads and rotary position embeddings. Input and output embeddings are tied, and the vocabulary has 50,257 tokens. The card ranks it level with its base model and behind SmolLM2 360M and Qwen2.5 0.5B on standard tests.

Write a User turn
Model continues as the Assistant
It ends with its end-of-answer token
A short reply comes back

Good to know:

• License: Apache 2.0, and the card asks you to check the data licences before redistributing derived data
• Cost to run: none named
• Data it reaches: downloads files from Hugging Face, then runs on your own machine
• Not for: added knowledge or reasoning, which the card says fine-tuning did not provide
• Limits: the card ranks it below larger peers, so expect weaker answers

On Hugging Face, 13 November 2026

• Task: text generation
• Library: transformers
• Downloads in the last 30 days: 0
• Likes: 8
• License: apache-2.0
• Languages: en
• Tags: transformers, text-generation, causal-lm, chat, gpt-neox, small-language-model, mine-o2, model-index

Links

• Model: https://huggingface.co/MineAITechnology/mine-o2-small
• Files: https://huggingface.co/MineAITechnology/mine-o2-small/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download MineAITechnology/mine-o2-small

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