| --- |
| title: README |
| emoji: 📊 |
| colorFrom: red |
| colorTo: green |
| sdk: static |
| pinned: false |
| --- |
| |
| <p align="center"> |
| <img src="https://cdn-uploads.huggingface.co/production/uploads/6921fa6332f7fb129563d495/aR36SrpWzksbcGbcp84pE.png" width="160"> |
| </p> |
|
|
| <h1 align="center">Jally AI by Lazy Loop Studio</h1> |
|
|
| <p align="center"> |
| Small, fast & indie‑built language models for real people, not just benchmarks. |
| </p> |
|
|
| --- |
|
|
| ## 🌱 Our Models |
|
|
| ### Nomi |
| Nomi is a series of **compact, high‑quality language models** designed to run on everyday hardware while still feeling surprisingly capable. |
| Instead of relying on massive data centers, Nomi focuses on efficient 3B models that you can run locally, privately, and with minimal setup. |
|
|
| Our models are primarily optimized for **clean structure**, **solid Markdown**, and **useful code**, making them a great fit for developers, indie projects, and anyone who wants full control over their AI stack. |
|
|
| ### Newest Models: |
| - **Nomi‑2** – Now with Qwen 3.5 as base. Improved math, code and token efficiency. |
|
|
|
|
| 👉 https://huggingface.co/JallyAI/Nomi-2 |
|
|
| --- |
|
|
|
|
| ## 🧩 Indie Spirit |
|
|
| Nomi AI is a small, independent project driven by a passion for **small but strong models**, not by a corporate budget. |
| We experiment with modern base models, fine‑tune them on carefully selected datasets, and try to squeeze as much “brain” as possible out of just 3B parameters. |
|
|
| If you care about local, open, and hackable AI tools, you are in the right place. |
|
|
| --- |
|
|
| ## 🔧 What We Build |
|
|
| - 🧠 3B models that feel like larger models (e.g., Nomi‑1.0, Nomi‑1.1). |
| - 📝 Assistants focused on **structured reports**, **Markdown**, and **developer workflows**. |
| - 🌍 Models with strong **German/English support** for European developer setups. |
|
|
| Whenever possible, our models are released under the **Apache‑2.0 license**, so you can freely use, modify, and integrate them into your own projects. |
|
|
| --- |
|
|
|
|
| ## 🗒️ Our Plan |
|
|
| Right now, we are fully focused on Nomi. Our goal is to develop the series up to Version 3 and make it competitive with much larger models: |
|
|
| - Nomi 1.0 → Basic instruction tuning |
| - Nomi 1.1 → Improved behavior plus some code and math |
| - Nomi 2 → Strong base, token efficiency. Great at math and code |
| - One step further → Agentic, tool‑using tasks that can stand alongside bigger models |
|
|
|
|
| --- |
| ## 🤝 Join the Journey |
|
|
| We welcome feedback, benchmarks, pull requests, and experiments using Nomi in your own projects. |
| If you are building on laptops, mini PCs, or homelabs and want something fast and lightweight, give Nomi a try and tell us what you built with it. |
|
|
| Check out our homepage (created by Nomi 2): [https://nomi-ai.netlify.app/](https://nomi-ai.netlify.app/) |
|
|
