OpenAI has made the Decisions API generally available.
The big shift is that developers now have a production-ready, low-latency path for bounded tasks such as routing, classification, and structured decision-making. OpenAI says the API is built with GPT-6 Luna and can run these workflows up to 10x faster than the Responses API for comparable use cases. That matters for agents and apps where small decisions add up across large volumes.
The practical takeaway is that teams can now send native JSON, include images in question instructions, and batch up to 400 questions per request. OpenAI also says prompt caching is supported for reused context, and the API includes native GPT-Live integration plus US and EU inference residency. Availability is immediate through the OpenAI platform. Developers building agentic products should evaluate whether to move repetitive decision steps into this API to cut latency and orchestration complexity.