Spaces:
Running
Running
| title: README | |
| emoji: π | |
| colorFrom: indigo | |
| colorTo: purple | |
| sdk: static | |
| pinned: false | |
| # π Novi AI | |
|  | |
| Novi AI is an experimental AI project focused on building **small, efficient language models**. | |
| ## π§ Models | |
| | Model | Parameters | Training Tokens | Status | | |
| | ------------- | ---------: | --------------: | ---------- | | |
| | **Novi-Nano** | ~1M | 300M | β Released | | |
| | Novi-Micro | 5M | 1B | π Soon | | |
| | Novi-Mini | 10M | 15B | π Soon | | |
| | Novi-Medium | 25M | 25B | π Soon | | |
| | Novi-Pro | 50M | 65B | π Soon | | |
| | Novi-Ultimate | 100M | 75B | π Soon | | |
| ### β Novi-Nano | |
| Novi-Nano is the **first released Novi model**, built as a compact language model for experimentation and lightweight AI. | |
| * π€ 8,192-token vocabulary | |
| * π FineWeb-Edu | |
| * π FineWeb-HQ | |
| * π SmolLM-Cosmopedia | |
| * β‘ GPU-accelerated training | |
| * π» Designed for efficient use | |
| ## π§ͺ Experimental Projects | |
| * **Novi-Code** β Compact text + code model | |
| * **Yuki** β Image-generation research | |
| * **Super-Novi** β Intentionally chaotic experiment | |
| * **Hyper-Novi** β Experimental model exploring unconventional training data | |
| These projects are experimental and may change, be renamed, or never be released. | |
| ## π Project History | |
| ### AppleMind | |
| Before Novi, there was **AppleMind** β an earlier experimental language-model project. | |
| AppleMind is now **retired**, with Novi becoming the primary project. | |
| **AppleMind β Novi AI β Novi-Nano** π | |
| --- | |
| *Small models. Big experiments.* | |