Fine-tuning is the process of taking a pre-trained model and continuing its training on a smaller, domain-specific dataset. It's the bridge between a general-purpose model and one that excels at your exact task.

When to Fine-Tune vs. Prompt Engineer

Prompt engineering is your first tool — it's fast, cheap, and requires no infrastructure. But it has limits:

• Consistent formatting: If you need every response in a precise JSON schema or specific tone, fine-tuning bakes this into the model's weights rather than relying on instructions.
• Domain knowledge: Medical, legal, or highly technical domains where the base model lacks depth benefit enormously from fine-tuning on curated examples.
• Latency & cost: Fine-tuned models can produce correct outputs with shorter prompts, reducing token costs and response time.
• Proprietary style: If you need outputs that match a specific brand voice across thousands of interactions, fine-tuning learns patterns that prompts can only approximate.

The Fine-Tuning Spectrum

Not all fine-tuning is equal. The field has evolved from full fine-tuning (updating every parameter) to increasingly efficient methods:

Key Terminology

• Base model: The pre-trained model you start from (e.g., Llama 3, Mistral, Gemma)
• Adapter: Small trainable matrices added to the frozen base model
• Epoch: One complete pass through your training dataset
• Loss: A number measuring how wrong the model's predictions are — lower is better
• Overfitting: When the model memorizes training data instead of learning patterns