OpenAI's Decisions API reaches GA and starts billing for caching
The question-answering endpoint leaves beta with double the question limit per request, but cache writes that were free in beta now cost standard rates.
OpenAI moved its Decisions API to general availability on 22 October, two and a half weeks after the public beta opened. The endpoint still runs only on gpt-6-luna and still returns one of three typed answers per question rather than free text, but a single request can now hold up to 400 questions, double the beta’s limit.
Two things changed with GA. Input to the shared input_json field can now include images, not just the per-question instructions. And caching pricing has moved to gpt-6-luna’s standard rates, with discounted cache reads but charged cache writes, where the beta had no caching charges at all.
{
"model": "gpt-6-luna",
"input": "shared evidence as text or message array",
"questions": [
{ "type": "choice", "name": "department", "instructions": "..." }
]
}Uncached input stays at $0.10 per million tokens, and output tokens are still free, which is the endpoint’s main pitch for routing and classification work.
Why it matters: anyone who built on the beta assuming cache writes were free needs to re-check their cost model before it ships. The doubled question limit is a straightforward win for batching classification or moderation calls into fewer requests.