litgpt vs torchtune
LLM training and fine-tuning toolkits compared
litgpt and torchtune are Python projects for training and adapting large language models with inspectable workflows and configurations. litgpt spans pretraining, fine-tuning, evaluation, and deployment, while torchtune focuses on post-training workflows and is no longer actively maintained.

litgpt: Train, Fine-Tune, and Deploy LLMs
LitGPT provides from-scratch implementations and training recipes for 20+ large language models, powered by Lightning Fabric from Lightning AI. Use its Python tools to pretrain, fine-tune, evaluate, and deploy models across single- or multi-device setups.

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 It is aimed at developers who want editable recipes and model-specific configs, but is no longer actively maintained.
| litgpt | torchtune | |
|---|---|---|
| Language | Python | Python |
| License | Apache-2.0 | BSD-3-Clause |
| Stars | 13.7k | 5.8k |
| Forks | 1.5k | 756 |
| Last analyzed | Nov 11, 2026 | Oct 3, 2026 |
Key differences
- litgpt includes from-scratch implementations and recipes for 20+ model families, while torchtune provides PyTorch implementations and configurations for models including Llama, Gemma, Mistral, Phi, and Qwen.
- litgpt covers pretraining, continued pretraining, fine-tuning, evaluation, and serving; torchtune focuses on post-training methods such as supervised fine-tuning, preference optimization, and knowledge distillation.
- litgpt documents HTTP inference serving and support for GPUs and TPUs; torchtune documents single-device, multi-device, and, for some recipes, multi-node training.
- litgpt uses the Apache-2.0 license, while torchtune uses BSD-3-Clause.
- litgpt is active, with recent commits and releases; torchtune development wound down in 2025 and is no longer actively maintained.
Choose litgpt if you…
- need workflows spanning pretraining, fine-tuning, evaluation, and serving.
- want from-scratch implementations and recipes across 20+ model families.
- need documented TPU support or HTTP inference serving.
Choose torchtune if you…
- want editable PyTorch recipes for post-training methods such as DPO or knowledge distillation.
- need YAML configurations and model-specific training workflows.
- are maintaining or studying an existing torchtune workflow and can plan for limited upstream support.
Compare the numbers side by side, or add more projects
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.