---
title: "token cost optimisation | Skills We Assign For | Azendo"
description: "Reducing LLM token cost — routing, caching, prompt size, plus the Azendo roles assigned for it."
url: "https://azendo.co/skills/token-cost-optimisation/"
---

[Skills](https://azendo.co/skills/) AI and machine learning 

# token cost optimisation.

Token cost optimisation reduces the spend of a language model system through model selection, prompt reduction, caching, batching and routing, without degrading the output quality that matters.

## Where token cost optimisation fits on a long engagement.

Routing by task is usually the largest available saving. Most systems send every request to their most capable model, while a substantial proportion of those requests — classification, routing, simple extraction — are handled identically by a model costing a fraction as much.

Prompt caching is the second lever and is frequently unused. Systems with a large fixed preamble pay for those tokens on every call, and providers offering caching reduce that substantially for no quality change. It is close to free money for anyone with a long system prompt.

## What an assigned team does with token cost optimisation.

Cost work needs quality measurement beside it, or it becomes quality reduction with a cost chart. A cheaper model that is adequate on most cases and poor on the ones that matter is not a saving.

Running both together is why cost work belongs with the evaluation harness rather than separately, agreed as part of [how the monthly fee is built](https://azendo.co/pricing/).

## What we use token cost optimisation for.

* Requests routed by difficulty Simple tasks handled by a cheaper model, which is usually the largest saving available.
* Fixed preambles cached Repeated context billed once rather than on every call.
* Quality measured alongside cost Evaluation run with every reduction, so savings are not quality loss.

## How token cost optimisation capacity is assigned.

Cost optimisation is assigned inside AI capacity alongside the evaluation harness, because a saving without a quality measure is not a saving.

## Roles we assign token cost optimisation for

* [LLM Engineer AI engineering](https://azendo.co/services/ai-engineering/llm-engineer/)

## Service lines it sits in

* [AI engineering](https://azendo.co/services/ai-engineering/)

Capacity is agreed as a committed monthly capacity across a discipline, not per skill.

## Related in ai and machine learning

* [vLLM — skill we assign for](https://azendo.co/skills/vllm/)
* [MLflow — skill we assign for](https://azendo.co/skills/mlflow/)
* [Kubeflow — skill we assign for](https://azendo.co/skills/kubeflow/)
* [OpenAI — skill we assign for](https://azendo.co/skills/openai/)
* [Anthropic — skill we assign for](https://azendo.co/skills/anthropic/)
* [Google Gemini — skill we assign for](https://azendo.co/skills/google-gemini/)
* [Meta Llama — skill we assign for](https://azendo.co/skills/meta-llama/)
* [Google and open-weight models — skill we assign for](https://azendo.co/skills/google-and-open-weight-models/)

## Tell us what your roadmap needs token cost optimisation for.

A service delivery manager replies with the disciplines we would assign, the monthly capacity and what the first month looks like.

[All skills we assign for](https://azendo.co/skills/)