---
title: README
emoji: 🚀
colorFrom: green
colorTo: blue
sdk: static
pinned: false
thumbnail: >-
https://cdn-uploads.huggingface.co/production/uploads/6851eab0c029515b3f067346/fSUgWp8DCv4FFufku7XNw.png
short_description: Open-Weights Very Small LLM Research
---
Welcome to Minty!
# What's Minty?
Minty is a non-profit organization that makes very small base and instruction-tuned LLMs from scratch.
Our mission is to train models that can run on a wide range of hardware (including CPUs)!
Currently, our models are trained on Google Colab, using NVIDIA T4 GPUs!
## Future Plans
| Model | Number of Parameters (and tokens, if applicable) | Variant | Will be Released on Minty Chat? |
|---------|-------------|-------------|-------------|
| Minty 1 Haiku | ~50M Parameters, ~3–5B Tokens | Base Model | No |
| Minty 1 Haiku Instruct | ~50M Parameters | Instruction-tuned Model | Yes |
## Other Future Plans
### Minty Chat
A chat platform, powered by Minty models. You ask questions, it generates responses!
### Planned Features
* ChatGPT-like UI
* *(Very future)* Model switcher
* Web searching (so it can access up-to-date information)
* *(Future)* Thinking mode
# Credits
* Logo: Generated by Design.com AI (picked the best one).
* Banner Design: Generated by GPT-5.6 Luna (and some minor edits by myself).
* Readme Card: Written by myself and then rewritten with GPT-5.6 Luna.
*by Mihai Popa, 2026*