OpenAI ships a Decisions API for routing, not chatting
A new endpoint takes a question with a fixed set of possible answers and returns a typed result, at a tenth of a cent per million input tokens with no output charge.
OpenAI opened its Decisions API to public beta on 6 October, a dedicated endpoint for the narrow job of picking an answer from a fixed set of options rather than generating free text. It runs on gpt-6-luna only, takes text or image input, and returns one of three typed answer shapes per question: a probability for a yes/no predicate, a confidence-scored choice from a fixed list, or a score against an ordered range.
The endpoint is POST /v1/decisions. A request names the model, the input, and an array of questions, each with a type, name and instructions:
curl https://api.openai.com/v1/decisions \
-d '{
"model": "gpt-6-luna",
"input": "I was charged twice for my order.",
"questions": [{
"type": "choice",
"name": "department",
"instructions": "Which department should handle this complaint?",
"choices": [{"value": "billing", "description": "Payments, invoices, and refunds."}]
}]
}'OpenAI says it returns answers roughly 10x faster than calling the same model through the Responses API. Pricing is $0.10 per million input tokens, with no charge for output, cache reads or cache writes.
Why it matters: routing, moderation and classification calls are usually the cheapest and most latency-sensitive part of an LLM pipeline, and a lot of people have been overpaying for a chat-completion call to do that job. A purpose-built endpoint with free output tokens changes the maths for anything that currently burns a full completion just to get back one of three labels.
The caveat: it’s public beta, restricted to one model, and OpenAI says general availability is still weeks away.