Glossary
Fine-tuning
Further training a base model on your own examples, to fix style, format or narrow task behaviour.
Fine-tuning adjusts a model's weights using your examples. It is effective for consistent output format, domain style, and narrow high-volume tasks where a smaller tuned model can replace a larger general one.
Data quality dominates. A few thousand carefully curated examples routinely beat tens of thousands of scraped ones, and the curation is most of the work.
Budget for maintenance. A fine-tuned model is a frozen artefact, as base models improve, your tuned version does not, and periodically re-tuning on a newer base is part of the running cost rather than an optional upgrade.
Commonly misunderstood: Most teams asking for fine-tuning need better prompting and retrieval instead. Fine-tuning genuinely wins on cost at volume, not usually on capability.
Related terms, in context
The concepts you almost always meet alongside fine-tuning.
- LoRA
- A fine-tuning method that trains a small set of extra parameters instead of the whole model.
- RAG
- Retrieving relevant passages from your own documents and giving them to the model, so answers are grounded and citable.
- Evaluation
- A labelled test set that turns 'it seems good' into a number you can regress against.
Where this shows up in our work
Fine-tuning is not an abstraction for us. It is a decision we make on live projects. It shows up most directly in custom model fine-tuning, where getting it wrong has a cost someone can measure.
If you are evaluating a vendor on this, the useful question is not whether they can define the term. It is what they measure, what they would refuse to do, and what happens in their system when the assumption behind fine-tuning stops holding.
Questions
What is Fine-tuning?
Further training a base model on your own examples, to fix style, format or narrow task behaviour.
What do people get wrong about fine-tuning?
Most teams asking for fine-tuning need better prompting and retrieval instead. Fine-tuning genuinely wins on cost at volume, not usually on capability.
Does Orqent Labs build this?
Yes, Custom Model Fine-tuning. We work across India, covering all 19,238 PIN codes remotely.
Building something that involves fine-tuning?
We will tell you honestly whether it is the right approach for your problem.
Or email bd@dtrasglobal.com · call +91 74118 77878