LLaMA-Factory

What it is for

LLaMA-Factory is an Apache-2.0 fine-tune UI: pick a base model, a dataset, LoRA or full weights, then train and export from LlamaBoard or the CLI. Use it when the job is fine-tune an open model on your GPU, not when you need a hosted chat seat.

What makes it unique: the cover is the official hiyouga/LLaMA-Factory GitHub file list — src/, examples/, Apache-2.0 in the About pane — not a wordmark.

Key features

• LlamaBoard — the Web UI for data, train, eval, and chat (README; llamafactory-cli webui). A console, not a script you paste.
• LoRA / QLoRA / full — one training job with adapters or full weights (README training methods). The train, not the UI.
• Multimodal — vision-language models as first-class datasets (README multimodal). Images in, not text-only SFT.
• Alignment — DPO, KTO, ORPO, PPO and friends on the same trainer (README). A preference job, not SFT.
• Export / merge / quantize — merge LoRA, GGUF / GPTQ export for local serve (README export). A package, not a train.

pip install / Docker / CLI are install / SDK. You bring the GPU bill — that is the meter. Base-model names are SKU.

Who it's for

Best fit: ML engineers who already have a GPU and will write a dataset, not a team looking for a $20 chat plan.

Pricing

Verified 2026-10-22 (Asia/Shanghai) from the Apache-2.0 LICENSE on the repo cover. There is no hosted SKU table.

• Software — Apache 2.0. License fee $0.
• GPU / API bill — you pay the cloud or the electricity. LLaMA-Factory does not print those dollars.

Do not invent a LLaMA-Factory Cloud seat.

Limitations / costs

• You bring the GPU. The UI is $0; A100 hours are not.
• Code-and-GPU first. There is no seat table. For hosted adapters without a cluster, see Unsloth or a cloud notebook.
• Fine-tune UI, not a prompt CMS. For versioned prompts, see PromptLayer.

Alternatives

• Unsloth — faster LoRA kernels when you already live in notebooks.
• Hugging Face — AutoTrain / TRL when you want their hosted path.
• DSPy — compile prompts, not weights.

Official tutorial

We searched for a first-party LLaMA-Factory YouTube getting-started lesson we could oEmbed. Community Colab walkthroughs are not embedded. Use written README – Getting Started (llamafactory-cli webui).

Website

https://github.com/hiyouga/LLaMA-Factory

Repository

https://github.com/hiyouga/LLaMA-Factory

Source

Catalog row curated for AI Tools Nav. Cover is the official hiyouga/LLaMA-Factory GitHub file list (Apache 2.0). License fee $0; no hosted seat table. No first-party YouTube embed.

Last verified

2026-10-22 (Asia/Shanghai) editorial check.