Fine-tuning
- Definition
- The process of continuing to train a pre-trained model on a smaller, task-specific dataset. Fine-tuning customizes model behavior for specific domains or formats and is a key part of most enterprise AI deployments.
- Why it matters
- Fine-tuning is how you turn a generic foundation model into your domain-specific competitive advantage. A general model knows a little about everything; a fine-tuned model knows a lot about your specific domain. For enterprises, fine-tuning lets you encode proprietary knowledge, enforce output formats, and achieve consistency that prompting alone cannot deliver. But fine-tuning is not free: it requires curated training data, ML engineering expertise, and ongoing maintenance as base models are updated. The decision of when to fine-tune versus when to use prompting and RAG is one of the most important architectural decisions in enterprise AI.
- In practice
- OpenAI's fine-tuning API charges $25/M training tokens for GPT-4o, making custom model training accessible without infrastructure. Companies like Harvey (legal), Abridge (healthcare), and Writer (enterprise content) fine-tune foundation models on domain-specific data and report 30-50% quality improvements over prompting alone. The open-source community uses LoRA and QLoRA to fine-tune Llama and Mistral models on a single GPU. A typical enterprise fine-tuning workflow involves: collecting 1,000-10,000 high-quality examples, splitting into train/eval sets, fine-tuning with LoRA, evaluating against a custom benchmark, and deploying behind an A/B test.
We cover data & training every week.
Get the 5 AI stories that matter — free, every Friday.
Related terms
LoRA (Low-Rank Adaptation)
A parameter-efficient fine-tuning technique that injects small trainable matrices into a frozen model. LoRA lets companies customize large models at a fraction of the cost of full fine-tuning.
Pre-training
The initial phase of model training where the network learns general knowledge from a massive dataset. Pre-training is the most expensive phase, often costing tens or hundreds of millions of dollars for frontier models.
Transfer learning
Using knowledge learned from one task or dataset to improve performance on a different but related task. Transfer learning is why pre-trained foundation models can be fine-tuned for specialized applications.
SFT (Supervised Fine-Tuning)
The process of training a pre-trained model on a curated dataset of input-output examples that demonstrate the desired behavior. SFT is typically the first alignment step after pre-training, teaching the model to follow instructions and produce useful responses.
Know the terms. Know the moves.
Get the 5 AI stories that matter every Friday — free.
Free forever. No spam.