Getting Started
Training Guides
Train larger models on less hardware. From free Kaggle GPUs to enterprise Blackwell.
Train 13B FREE on Kaggle
FreeLlama-2 13B on free Kaggle T4. 17.3x compression. $1,500 → $0.
70B on Single GPU
EnterpriseLlama-2 70B on Blackwell. 15.7x compression. 840GB → 53GB.
Calculate Your Savings
ToolSee how much memory, cost, and energy you can save with AXIOM.
Just Want to Get Started?
Three lines of code. Works with any PyTorch or HuggingFace model.
$ pip install quarterbit