Fine-tuning

Full definition

Modern fine-tuning is usually parameter-efficient — methods such as LoRA (Low-Rank Adaptation) and QLoRA update only a small fraction of weights, making fine-tuning feasible on a single GPU. Full fine-tuning updates all weights and requires significant compute. Fine-tuning is most justified when the task has a distinctive style or narrow distribution that prompting + RAG cannot capture; otherwise prompting is usually faster and cheaper.

Why it matters

Fine-tuning is now a real procurement decision: build a custom model on top of an open-weights base (Llama, Mistral, Qwen) and own the artifact, or pay an API per token forever. For high-volume or sensitive workloads, fine-tuned models on owned infrastructure can deliver 60-90% lower TCO than equivalent API usage.

Example

A law firm fine-tunes Llama 3 70B on 5 years of redacted internal memos, producing a private model that drafts in the firm's house style — something prompt engineering alone could not reliably achieve.

Related terms

• Reinforcement Learning from Human Feedback (RLHF)
  Reinforcement Learning from Human Feedback (RLHF) is a training technique that aligns a language model's outputs with human preferences by training a reward model on human comparisons of candidate responses, then optimizing the LLM against that reward.
• Pretraining
  Pretraining is the first, most compute-intensive stage of training a foundation model — typically self-supervised next-token prediction over trillions of tokens of text, code, and other modalities — that produces the base model later fine-tuned for specific tasks.
• Transfer Learning
  Transfer learning is the practice of using a model trained on one task as the starting point for a related task, dramatically reducing the data and compute required to reach high accuracy on the new task.
• Foundation Model
  A foundation model is a large model trained on broad data at scale — typically self-supervised — that can be adapted to many downstream tasks via prompting, fine-tuning, or retrieval, instead of being trained task-by-task.
• Prompt Engineering
  Prompt engineering is the practice of designing the inputs to a language model — instructions, examples, role definitions, output formats, and constraints — to reliably produce a desired output without changing the model's weights.