OpenAI Decisions API generally available for fast classify-and-route calls
OpenAI says the Decisions API is generally available on gpt-6-luna, returning predicate probabilities, choices, or scores over text, images, and JSON about 10x faster than the Responses API. Since beta it adds prompt caching, native JSON, up to 400 questions per request, and GPT-Live integration. The openai.com post returned HTTP 403; this piece follows the API docs and changelog only.
OpenAI’s API docs state that the Decisions API is now generally available. It targets the many bounded decisions apps and agents make—classifying inputs, routing requests, or picking the next action—and returns typed answers over text, images, and JSON: a predicate probability, a choice from a fixed set, or a score against ordered levels. OpenAI says it is about 10x faster than the Responses API. Requests go to POST /v1/decisions; the only supported model is currently gpt-6-luna.
Since public beta, GA adds prompt caching with cache discounts, images in question instructions, native input_json, up to 400 questions per request (up from 200), hosted image URLs, conversation history with tool calls, organization usage reporting, native GPT-Live integration, and Private Link. The API changelog dates the GA update to Oct 22, 2026.
Note: capture of openai.com/index/decisions-api returned HTTP 403 and was not bypassed. Facts here come only from the reachable developers.openai.com guide and changelog. Customer quotes or residency claims that appear only on the blocked blog post are omitted.
Sources
- Decisions | OpenAI API — OpenAI
- Changelog | OpenAI API — OpenAI