What it does
Core capabilities at a glance
- YAML-based configuration - no coding needed for training
- Supports full fine-tuning, LoRA, QLoRA, and more
- Multi-GPU and multi-node training
- 100+ model architecture support
- WANDB and MLFlow experiment tracking
- DeepSpeed and FSDP integration
Deep dive
The full breakdown - performance, comparisons, and setup
Axolotl
Axolotl is the fine-tuning framework you reach for when you need control and reproducibility. Instead of writing custom training loops, you define your training run in a YAML file, and Axolotl handles the rest.
What it is
Axolotl is a configuration-driven fine-tuning framework that supports the widest range of model architectures among open-source training tools. It wraps HuggingFace Transformers with optimizations for memory efficiency, multi-GPU training, and experiment tracking.
How it stacks up
| Axolotl | Unsloth | LLaMA-Factory | LitGPT | |
|---|---|---|---|---|
| Config format | YAML | Python | Web UI | Python |
| Model coverage | 100+ | 20+ | 100+ | 20+ |
| Training methods | All | LoRA/QLoRA | All | Full/LoRA |
| Multi-GPU | ✓ | ✓ | ✓ | ✓ |
| Best for | Research, control | Speed, memory | Web UI users | Full pretraining |
Get started
# config.yml
model:
name: Qwen/Qwen3-8B
load_in_8bit: true
training:
learning_rate: 2e-4
num_epochs: 3
batch_size: 4
micro_batch_size: 2
lora:
r: 16
alpha: 16
dropout: 0.05accelerate launch -m axolotl.cli.train config.ymlWhen to use something else
- Speed and memory optimization: Unsloth
- Web UI interface: LLaMA-Factory
- Full pretraining: LitGPT or Torchtune
From the community
What users are saying
“Axolotl's have become a global icon. First as an anti-colonial protest symbol for indigenous peoples. But now it's even a creature in Minecraft Edit: oh the article says as much > Axolotls as pets have seen a surge in popularity in recent years after they were introduced to video games such as Minecraft and Roblox.”
“The range used to be somewhat larger, but the environment has been reduced (most of what’s now Mexico city used to be a lake).”
“> An Axlotl tank or Axolotl tank was the "tank" by which the Bene Tleilax reproduced a living human being from the cells of a cadaver, producing type of clone called a ghola.”
Frequently asked
Quick answers to common questions
What is Axolotl?
Axolotl is a fine-tuning tool for local AI workloads. Streamlined fine-tuning framework for LLMs with YAML configs, supporting 100+ model architectures.
Is Axolotl free and open source?
Yes, Axolotl has 12,591 GitHub stars and is licensed under Apache-2.0. You can self-host it for free on linux, docker.
What platforms does Axolotl support?
Axolotl runs on linux, docker.
What hardware do I need for Axolotl?
The hardware requirements depend on which models you run. Check our hardware directory for compatible GPUs and systems. Axolotl has 12,591 GitHub stars and an active community.
Does Axolotl support GPU acceleration?
Axolotl supports GPU acceleration via CUDA, Metal, or Vulkan depending on your platform. For the best performance, pair it with an NVIDIA RTX 4090 or 5090.
What are the best alternatives to Axolotl?
Popular alternatives include other fine-tuning tools in our directory. Browse our full collection at /tool for comparisons, community reviews, and benchmark data to find the right fit for your workflow.
How much does Axolotl cost?
Axolotl is free-open-source. It is completely free and open source to self-host.
Pairs well with
Complementary tools, models, and hardware