Model components / Fine-tuning code

LLaMA-Factory

Unknown

LLaMA-Factory (hiyouga/LLaMA-Factory), Apache-2.0, confirmed live June 2026 (~71.9K stars). Unified efficient fine-tuning of 100+ LLMs/VLMs (ACL 2024); CLI + web UI; LoRA/QLoRA/full FT and preference/RL methods.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public(hiyouga/LLaMA-Factory)
core-gated: ungated

Fully OSI-licensed (Apache-2.0), full source public including the web UI; no feature-gated core.

https://github.com/hiyouga/LLaMA-Factory recorded 2026-06-04
Apache-2.0 license, unified fine-tuning framework

Adoption

3 medium confidence

3.0

~234.6K PyPI (llamafactory) downloads last month plus very large GitHub adoption (~71.9K stars, 8.8K forks) and named users (Amazon, NVIDIA, Alibaba/Aliyun). Many users run from source/WebUI rather than pip, so pip undercounts; level 3 on download volume + named-deployment breadth, stars corroborate only.

https://pypistats.org/packages/llamafactory recorded 2026-06-04
Downloads last month: 234,631
https://github.com/hiyouga/LLaMA-Factory recorded 2026-06-04
~71.9K stars; named users Amazon/NVIDIA/Alibaba/Aliyun

Capability

4 medium confidence

4.0

Strong breadth (100+ models, VLMs, no-code UI) puts it at C4; lacks the demonstrated frontier-scale parallelism of the Megatron-LM C5 anchor.

https://github.com/hiyouga/LLaMA-Factory recorded 2026-06-04
100+ LLMs/VLMs, LoRA/QLoRA/full FT, CLI+WebUI

Unchanged since 2026-07-30 (last edited, not re-checked)