# xTuring

A library for fine-tuning, evaluating, and running personalized, private open-source large language models locally or in a private cloud.

xTuring makes it simple, fast, and cost-efficient to fine-tune open-source large language models such as LLaMA, LLaMA 2, Qwen3, GPT-OSS, and GPT-J on a user's own data, running locally or inside their own private cloud rather than through a hosted provider. A simple API covers data preparation, training, and inference in a few lines of code.

It supports LoRA and low-precision INT8/INT4 training to cut compute cost, scales from a CPU or a laptop up to multiple GPUs, and includes built-in evaluation metrics such as perplexity. Recent additions include off-the-shelf, LoRA, and quantized configurations for large GPT-OSS models with configurable reasoning levels, and support for LLaMA 2 through a generic model interface.

It is written in Python, requires a minimum recent version of the transformers library for its INT8/INT4 loading code, and is released under the Apache License 2.0.

Website: https://xturing.stochastic.ai