Skip to content
repository radar
Models & Inference

🦙LLaMA-Factory

ACTIVE STEADY

hiyouga/LlamaFactory · homepage ↗

Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)

⭐ Hugely popular: 74k stars, gaining about 188 a week

View on GitHub ↗

repo profile

vintage 2023 3 years old
delivery library
language Python
license Apache-2.0

momentum

total stars 74k
stars added last week +188
weekly downloads 6k /wk · PyPI
used by 110 repos & packages
commits / week 4 cooling · -29% vs prior mo
issues closed 49% ~1 day to close, median
contributors 294 (+5 in 15d)
release cadence quarterly
last activity 2d ago

durability

backing community / independent
openness permissive
bus factor 5 distributed
top-author share 13% 6 mo

bus factor = how many people it takes to cover more than half the commits (6 months). 1 is a solo project; higher means the work is spread across a team. top-author share is the single busiest author's slice of those commits.

since we covered it

monthly average + 1k/mo (+2%/mo) · + 15k total since PR#18

why it's a big deal

  • Consolidates fine-tuning across virtually all major open-source models into one consistent interface.
  • Covers the full spectrum: pre-training, supervised fine-tuning, preference modeling, and RLHF variants.
  • Makes cutting-edge methods like LoRA, QLoRA, DoRA, OFT, and reward modeling usable without bespoke engineering.

under the hood

  • Supports over 100 LLMs and VLMs (LLaMA, Mistral, Qwen, DeepSeek, Gemma, Yi, Phi, etc.).
  • CLI and WebUI powered by Gradio with OpenAI-compatible inference endpoints.
  • Optimized for multiple backends (PyTorch, vLLM, DeepSpeed, FlashAttention-2, bitsandbytes).

our take from PR#18, 2025-10-01

star history

PR#18 · 60k74k now May 2023Aug 2026
  1. PR#18 60k 2025-10-01
  2. now 74k + 15k since first covered

curve is sampled from GitHub's star history, plus our own daily readings since we covered it; the dashed stretch is before we first covered it, the solid line since. figures at coverage are the numbers we printed then (approx.), current count is live.

understory

Better known than its recent output, coasting a little on attention.

-15 understory score output 57 · clout 72
Aug 2025 Jul 2026
  • output, commits & releases
  • clout, star velocity

output = commits & releases; clout = star velocity, both 0 to 100 monthly indices; the gap where output runs above clout is the understory. The understory →

covered in

  • PR#18 2025-10-01 on the radar

    Unified Fine-Tuning of 100+ LLMs & VLMs

similar projects

compare these →
  • Unsloth

    Local UI to run and train LLMs and diffusion models, including Qwen3.8, Kimi K3, MiniMax-H3, Gemma 4, DeepSeek-V4, FLUX and more.

    72k ACTIVE
  • 🧩 TensorZero

    Rust · leaner, 12k stars

    TensorZero is an open-source LLMOps platform that unifies an LLM gateway, observability, evaluation, optimization, and experimentation.

    12k OFF THE RADAR
  • 🧠 UltraRAG

    leaner, 6k stars

    A Low-Code MCP Framework for Building Complex and Innovative RAG Pipelines

    6k ACTIVE

comments

Sign in with GitHub to add your blip on LLaMA-Factory.

loading comments…