axolotl vs torchtune
LLM fine-tuning and post-training frameworks compared
axolotl and torchtune are Python projects for configuring and running LLM post-training workflows, including supervised fine-tuning and preference optimization. axolotl emphasizes a broad, YAML-configured workflow with multimodal and distributed options, while torchtune provides editable PyTorch recipes and model-specific configurations but is no longer actively maintained.
axolotl: Fine-Tune Large Language Models
Axolotl is a Python framework for fine-tuning and post-training language models through configurable workflows. It supports methods from LoRA and QLoRA to preference tuning and reinforcement learning, with options for multimodal and distributed training.
torchtune: Fine-Tune and Post-Train Large Language Models
torchtune is a PyTorch library for configuring and running LLM post-training workflows, from supervised fine-tuning to preference optimization. It is aimed at developers who want editable recipes and model-specific configs, but is no longer actively maintained.
| axolotl | torchtune | |
|---|---|---|
| Language | Python | Python |
| License | Apache-2.0 | BSD-3-Clause |
| Stars | 12.5k | 5.8k |
| Forks | 1.5k | 756 |
| Last analyzed | Oct 3, 2026 | Oct 3, 2026 |
Key differences
- axolotl covers supervised and preference tuning, reinforcement learning, and supported vision, audio, and video models; torchtune also includes knowledge distillation and packages recipes for specific model families.
- axolotl uses YAML configuration across preprocessing, training, evaluation, quantization, and inference; torchtune centers on inspectable PyTorch recipes and YAML configs that can be edited or overridden.
- axolotl lists distributed options including FSDP2, DeepSpeed, and multi-node workflows; torchtune supports single-device and multi-device training, with multi-node options for some recipes.
- axolotl is licensed under Apache-2.0; torchtune is licensed under BSD-3-Clause, while its third-party model weights may have separate terms.
- axolotl is not archived; torchtune's development wound down in 2025, so it may suit maintaining existing workflows better than projects that rely on ongoing upstream support.
Choose axolotl if you…
- need a configurable workflow spanning fine-tuning, evaluation, quantization, and inference.
- want to run supported multimodal training or use distributed integrations such as FSDP2 and DeepSpeed.
- plan to experiment with LoRA, QLoRA, preference tuning, or reinforcement learning in one framework.
Choose torchtune if you…
- want editable PyTorch recipes and model-specific configurations.
- need to inspect or adapt existing torchtune workflows, including knowledge distillation or preference optimization.
- want to investigate memory and performance techniques such as activation offloading or lower-precision optimizers.
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.