Axolotl

non-cncf

First Research Pass

GitHub Repository Official Website

Content Lifecycle & Evolution State

1

Initial Seed

Metadata Loaded

2

First Research

Automated Gathering

3

Tech Writing

Content Structured

4

Approval

Editorial Verified

First Research Pass Completed: Initial facts, key features, and documentation links have been gathered. Detailed technical writing and synthesis are currently in queue.

Overview

Axolotl is a free and open-source tool designed to streamline post-training and fine-tuning for the latest large language models (LLMs).

Key Features

Multiple Model Support: Train various models like GPT-OSS, LLaMA, Mistral, Mixtral, Pythia, and many more models available on the Hugging Face Hub.
Multimodal Training: Fine-tune vision-language models (VLMs) and audio models.
Training Methods: Full fine-tuning, LoRA, QLoRA, GPTQ, QAT, FP8 mixed-precision training, NVFP4/MXFP4 MoE LoRA, Preference Tuning, RL, and Reward Modelling.
Performance Optimizations: Multipacking, Flash Attention 2/3/4, Xformers, Flex Attention, Sequence Parallelism, Multi-GPU and Multi-node training, etc.
Easy Configuration: Re-use a single YAML configuration file across the full fine-tuning pipeline.

Use Cases

LLM Post-Training and Fine-Tuning. Developers can quickly fine-tune various state-of-the-art models for specific tasks using LoRA, QLoRA, and other methods with optimized performance and multimodal support.

Getting Started

```bash
# Fetch axolotl examples
axolotl fetch examples

# Train a model using LoRA
axolotl train examples/llama-3/lora-1b.yml

Recent Updates

Release v0.19.0 was published on September 10, 2026. Axolotl has recently added FP8 mixed precision training via torchao, TiledMLP support for single-GPU to multi-GPU training, Quantization Aware Training (QAT) support, Llama 4 support, and Sequence Parallelism (SP) support.

Did You Know?

Axolotl ships with built-in documentation optimized for AI coding agents (Claude Code, Cursor, Copilot, etc.), bundled directly with the pip package.