Custom Swift and Metal runtime enabling 26B instruction-tuned model inference in around 2 GB RAM by streaming weights from SSD, optimized for Apple Silicon Macs.
United Kingdom
A user-friendly interface built on Tinker API that lets you fine-tune LLMs, chat with your trained model, and deploy to Hugging Face.
Cost / License
- Free
- Open Source (Apache-2.0)
Platforms
- Windows
- Mac
- Linux
Best alternatives are1 alternatives
Italy
EU
