Fine-tuning is the process of further training a pre-trained AI model on a smaller, specialized dataset to adapt it to a specific task, domain, or style. It is one of three main ways to customize a foundation model, alongside prompting and retrieval-augmented generation.
How it works
You start from a model that already learned general language or vision from massive data, then continue training it on a curated set of examples for your use case. This adjusts the model’s weights so its outputs better match the target domain — a legal assistant, a brand voice, a coding style. Modern variants (such as LoRA) fine-tune only a small fraction of parameters to cut cost.
Why it matters
Fine-tuning is how organizations turn general models into specialized ones. Talent able to fine-tune and deploy models is a noted bottleneck to enterprise adoption — see Enterprise AI Statistics 2026.
Related terms: Foundation Model · RAG · LLM · All glossary entries