N環球 
OpenAI's Decisions API Goes Generally Available: 400 Questions Per Call, Latency Cut to Milliseconds
OpenAI moved its Decisions API from public beta to general availability on October 22: powered by GPT-6 Luna, it returns only typed answers — a probability, a choice or a score — with up to 400 questions per request, $0.10 per million input tokens and no output-token charge.
Why it matters
Chatbots write prose; the Decisions API only judges: approve the claim or not, which department gets the ticket, how severe is the bug — the small decisions apps make thousands of times a day can now be batched into one API call, with latency low enough for real-time workflows.
Full report
According to OpenAI's developer community announcement on October 22, the Decisions API is out of beta and generally available. It was previewed at DevDay on September 29 and opened to public beta on October 6, reaching general availability 23 days later.
The API is powered by GPT-6 Luna at the endpoint POST /v1/decisions. It returns no prose, only three typed answers: predicate (a probability from 0 to 1 that a condition holds), choice (one value from a fixed set, with a probability for every option and a confidence) and score (a rating against ordered levels). A single request can carry up to 400 questions, double the beta limit; new features include native GPT-Live voice integration, prompt caching, native JSON inputs and images in individual question instructions.
On price: $0.10 per million uncached input tokens, $0.01 per million cached input tokens, and no output-token charges. OpenAI API product lead Nikunj Handa said latency fell from 1–3 seconds to 100–180 milliseconds; early trial customers include language-learning app Speak, Box, Datadog and Ramp, with Speak reporting the gap between actions dropping from 1.4 seconds to 0.13 seconds. OpenAI says the service runs about 10 times faster than routing the same bounded decisions through its standard Responses API.
💬 Comments
No comments yet. Be the first!