maxtext

Overview

maxtext is a Jax-based framework for building, training, and scaling large language models. It emphasizes simplicity and performance while supporting distributed training across multiple accelerators.

Best for

Best for
Developers already using Jax who need a streamlined, scalable LLM training framework

Use cases

• Training large language models from scratch using Jax
• Scaling LLM training across TPUs or GPUs with minimal code changes
• Experimenting with model architectures and hyperparameters in Jax

Notes

maxtext is a Jax-based framework for building, training, and scaling large language models. It emphasizes simplicity and performance while supporting distributed training across multiple accelerators.

2,303 stars on GitHub. Last updated 2026-06-01. Licensed Apache-2.0.

Use cases

• Training large language models from scratch using Jax
• Scaling LLM training across TPUs or GPUs with minimal code changes
• Experimenting with model architectures and hyperparameters in Jax

Pros

• Designed for simplicity, reducing boilerplate compared to raw Jax
• Built for performance and scalability across hardware
• Open source with a growing community (2303 stars)

Cons

• Limited to Jax ecosystem, not compatible with PyTorch or TensorFlow
• Smaller community and fewer pre-built components than mainstream frameworks
• Documentation and examples may be less extensive than commercial alternatives

Indexed from awesome-llm and enriched against its public facts.