Petuh 2.0
Collection
2 items • Updated
🐔 Petuh-2.0
The fastest lightweight reasoning model trained from scratch on 4.72B tokens and fine-tuned on pure entropy.
Technical Specifications:
- Total Parameters: 108,943,104 (~108.9M)
- Vocabulary Size: 8,192
- Embedding Dimensions: 768
- Hidden Layers: 14
- Attention Heads: 12 (Query) / 4 (KV, GQA)
- Context Length: 1,536
Benchmark Results:
| Benchmark | Score |
|---|---|
| AIME 2025 | 0.0 |
| HLE | 0.0 |
| Terminal bench 2.1 | 0.0 |
| SWE Bench verified | 0.0 |
- Downloads last month
- 285
Hardware compatibility
Log In to add your hardware
3-bit
4-bit
5-bit
6-bit
8-bit
16-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support