MaxText
A high-performance, scalable LLM library and reference implementation written in pure Python/JAX for training on TPUs and GPUs.
What is MaxText?
MaxText is a high-performance, scalable library and reference implementation for training large language models (LLMs) written in pure Python/JAX. It provides a collection of state-of-the-art models and efficient training pipelines for both pre-training from scratch and post-training techniques like supervised fine-tuning and reinforcement learning. The library is designed to achieve maximum hardware utilization on TPUs and GPUs while maintaining a simple, optimization-free codebase.
AI researchers and engineers who need to train or fine-tune large language models at scale, particularly those using Google Cloud TPUs or high-performance GPU clusters. It's also suitable for teams building custom LLMs for production or research who want a performant, open-source foundation.
Developers choose MaxText for its exceptional performance and scalability out-of-the-box, thanks to JAX's XLA compiler optimizations. Unlike many LLM frameworks, it achieves high Model FLOPs Utilization without requiring manual low-level tuning, while providing a comprehensive model library and support for both pre-training and advanced post-training techniques.
Overview
A simple, performant and scalable Jax LLM!
Use Cases
Best For
Related Projects
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
