Home → Use cases → Support and chat bots

Support and chat bots

Support traffic is high-volume and repetitive, which makes it the workload where careless model choice is most expensive and careful prompt design pays back fastest.

Output length is the lever

Output tokens cost several times what input tokens cost on every model we carry — five times on the current Claude and GPT flagships. A support bot that answers in three sentences instead of three paragraphs costs a fraction of one that does not, with no change of model and usually a better customer experience.

This is worth saying plainly because it is the opposite of the usual instinct. When a bill is too high the reflex is to drop a tier and re-validate everything. Capping answer length in the prompt is faster, reversible, and often larger.

Split the routing from the answering

Most support systems classify before they answer: is this billing, technical, or a refund request; is it in scope; does it need a human. Those decisions are short, structured and high-volume, and a small model does them as well as a large one. gpt-5.6-luna is priced for exactly this shape of call.

Keep the flagship for the reply the customer reads. That single split usually moves the blended cost closer to the cheap model than the expensive one, because routine calls dominate the count.

Before you point it at customers

• Decide what happens when a model call fails. There is no cross-provider fallback here — a failed request fails and your code handles it. A support bot needs a human path, not a retry loop.
• Understand where the prompt goes. Requests pass through to the model provider and content is not written to our database; we keep metadata for billing. The provider serving the model receives the prompt and applies its own policy. If you are handling customer data under an agreement, read the direct-account comparison before deciding.
• Set a spend expectation. Balance is prepaid here, so the failure mode of a runaway loop is that calls stop, not that an invoice arrives. Watch the balance the first week.

Setup

Base URL https://apiclan.us/v1 for OpenAI-lineage clients and SDKs. Model names and prices are on the price list; each model has a page showing the official rate beside what you pay and a calculator for your own volume.

Start using it

No subscription, no monthly minimum, no sales call. Top up with USDT and spend what you use — 1 USDT gives you 2 credits of API balance.

Read the 30-second quickstart

Prices quoted on this page are regenerated automatically from live billing data. Third-party terms are quoted from that party's own published documentation.