Source: https://seerepo.com/compare/hiyouga/LlamaFactory/vs/ostris/ai-toolkit
Markdown source: https://seerepo.com/compare/hiyouga/LlamaFactory/vs/ostris/ai-toolkit.md
Description: LlamaFactory and ai-toolkit share 15 technologies: FastAPI, uvicorn, Gradio, matplotlib, einops, sentencepiece, transformers, accelerate, peft, safetensors, bitsandbytes, torchao, torchcodec, av, and Pydantic. Both are written in Python.

# LlamaFactory vs ai-toolkit

LlamaFactory and ai-toolkit share 15 technologies: FastAPI, uvicorn, Gradio, matplotlib, einops, sentencepiece, transformers, accelerate, peft, safetensors, bitsandbytes, torchao, torchcodec, av, and Pydantic. Both are written in Python.

- [hiyouga/LlamaFactory](https://seerepo.com/r/hiyouga/LlamaFactory): Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)
- [ostris/ai-toolkit](https://seerepo.com/r/ostris/ai-toolkit): The ultimate training toolkit for finetuning diffusion models

## What they share and what sets them apart

- In common: FastAPI, uvicorn, Gradio, matplotlib, einops, sentencepiece, transformers, accelerate, peft, safetensors, bitsandbytes, torchao, torchcodec, av, Pydantic
- Only in LlamaFactory: aqlm, eetq, gptqmodel, hqq, ktransformers, tiktoken, torchvision, triton-ascend, vector-quantize-pytorch, vocos, torch, datasets, trl, badam, deepspeed, torch-npu, vLLM, fsdp-turbo, liger-kernel, ModelScope, openmind, torchaudio, adam-mini, apollo-torch, galore-torch, nltk, optimum, rouge-chinese, sglang, torchdata, transformer-engine, jsonschema-specifications, referencing, swanlab
- Only in ai-toolkit: Hugging Face, React, Next.js, SQLite, node-cache, albucore, kornia, lycoris-lora, open-clip-torch, timm, optimum-quanto, opencv-python, controlnet-aux, albumentations, librosa, lpips, prodigyopt, pytorch-fid, invisible-watermark, mutagen, Tailwind CSS, classnames, Headless UI, react-dropzone, @monaco-editor/react, react-virtuoso, react-zoom-pan-pinch, react-select, uplot, Lucide, react-icons, Prisma, yaml

## Side by side

| | LlamaFactory | ai-toolkit |
|---|---|---|
| Size | 62,648 lines written, 194 code files | 204,162 lines written, 663 code files |
| Who writes it | 1.6% of its commits signed by AI. With Gemini CLI, Claude Code, GitHub Copilot, and Cursor | 0.2% of its commits signed by AI. With Codex, Cursor, and Claude Code |
| Activity | Last push November 9, 2026, 2,772 commits, 75,956 stars | Last push November 10, 2026, 1,578 commits, 12,609 stars |
| Tests | 62 test files. 38.1% of its code files have a test that imports them | 12 test files. 4.5% of its code files have a test that imports them |

- Size: Lines written by people and agents, without generated ones; code files, without tests.
- Who writes it: A commit counts as AI's when it carries its signature: a Co-Authored-By line, its author name or the merge of its branch.
- Activity: From GitHub and its history, on the day SeeRepo read it.
- Tests: The files that test the code, and the share of code files a test imports.

## Frequently asked questions

### Which is bigger, LlamaFactory or ai-toolkit?

ai-toolkit, with 204,162 lines written; LlamaFactory has 62,648.

The proof: [ostris/ai-toolkit's code at commit 4a7f285](https://github.com/ostris/ai-toolkit/tree/4a7f28532ed6fe81d9f8c00718878bb3b8739015)

### Which has more GitHub stars, LlamaFactory or ai-toolkit?

LlamaFactory, with 75,956 stars; ai-toolkit has 12,609.

The proof: [hiyouga/LlamaFactory on GitHub](https://github.com/hiyouga/LlamaFactory)

### Which has more tests, LlamaFactory or ai-toolkit?

LlamaFactory, with 62 test files; ai-toolkit has 12.

The proof: [hiyouga/LlamaFactory's code at commit 28462f1](https://github.com/hiyouga/LlamaFactory/tree/28462f15cc00b8233afab64c7712650d0a18a273)

### Which was updated more recently, LlamaFactory or ai-toolkit?

ai-toolkit: its last push was on November 10, 2026, and LlamaFactory's on November 9, 2026.

The proof: [ostris/ai-toolkit's commits on GitHub](https://github.com/ostris/ai-toolkit/commits)

### Which has more commits signed by AI, LlamaFactory or ai-toolkit?

LlamaFactory: an AI signed 1.6% of its commits, with Gemini CLI, Claude Code, GitHub Copilot, and Cursor; in ai-toolkit, 0.2%.

The proof: [the commits each tool signed in hiyouga/LlamaFactory](https://seerepo.com/r/hiyouga/LlamaFactory/ai)

## Sources

- hiyouga/LlamaFactory: Read on November 9, 2026, at commit 28462f1 (https://github.com/hiyouga/LlamaFactory/tree/28462f15cc00b8233afab64c7712650d0a18a273)
- ostris/ai-toolkit: Read on November 10, 2026, at commit 4a7f285 (https://github.com/ostris/ai-toolkit/tree/4a7f28532ed6fe81d9f8c00718878bb3b8739015)

## What is your repo built with?

Any public repo, with no account: https://seerepo.com/r/owner/repo/built, with the owner and name it has on GitHub.