Product descriptions at a twentieth of the cost

The situation

The business needed descriptions written for a higher-level aggregated data product. Proprietary model services could do it, but at a price that grows with every description.

What we did

We hosted an open-weight model inside the client’s own network, on compute that runs only when it is needed, and compared it with the proprietary services.

What changed

A twentieth of the price of proprietary endpoint services, with no loss of description quality. The data also never left the client’s network.

Why it matters

The most capable model is not always the right one. This is the principle in practice: minimise AI spend, and choose the model to suit the job.