Fine-tuning
Further training of a pretrained LLM on a smaller, domain-specific dataset to specialise its behaviour — for tone, format, vocabulary, or task accuracy. Cheaper than training from scratch but still requires curated data and compute.
Background
Fine-tuning takes a base model and updates its weights (full fine-tuning) or adds small adapter layers (LoRA, QLoRA) so it performs better on a target distribution. For vibe-coding tools, fine-tuning is rarely needed at the application layer — frontier models with good system prompts cover most use cases. Where it shows up: code-completion vendors fine-tuning smaller models for latency, enterprise customers fine-tuning on internal code style, and providers releasing coding-specialised checkpoints (DeepSeek-Coder, Codestral).