# What is fine-tuning?

Fine-tuning continues a model’s training on your own examples so it learns your domain, formats, and decision patterns. A fine-tuned open-weight model typically beats a much larger general model on your specific task, at a fraction of the per-token cost, and the resulting weights can belong to you.

The inputs are pairs of task and outcome: resolved claims, cleared disputes, accepted answers. A few thousand high-quality examples usually move accuracy more than any prompt engineering.

On Allocate, fine-tuning runs against open-weight bases like Qwen 3.5, Llama 4 Scout, and GLM. The trained weights stay inside your boundary and belong to you; that ownership is the difference between renting intelligence and building an asset.

## See also

- [Explore fine-tunable models](https://allocate.network/models)
- [Fine-tuning cost calculator](https://allocate.network/tools/fine-tuning-cost-calculator)

## Related terms

- [Open-weight models](https://allocate.network/glossary/open-weights.md)
- [Training signal](https://allocate.network/glossary/training-signal.md)
- [Model distillation](https://allocate.network/glossary/model-distillation.md)

---

[HTML page](https://allocate.network/glossary/fine-tuning) · [Machine-readable catalog](https://allocate.network/catalog.json)