OpenAI Launches Decisions API for Faster App Workflows
OpenAI's Decisions API is now generally available, powered by GPT-6 Luna to help apps and agents make rapid, structured choices.
Modern software applications and autonomous agents constantly execute routine micro-choices throughout their execution flows. Tasks like routing incoming requests, classifying user text, or choosing subsequent actions might look trivial in isolation, but their aggregate latency and execution expenses compound quickly at scale. To address this friction, OpenAI has announced the general availability of the Decisions API.
Driven by GPT-6 Luna, this endpoint targets bounded questions and operates up to ten times faster than standard queries through the Responses API. Since its public beta phase, the service has gained several enhancements, including a native GPT-Live integration, prompt caching for repeated context, and regional inference residency options spanning the United States and Europe. Furthermore, developers can submit native JSON inputs, attach images directly to instructions, and scale queries up to 400 questions per single request—double the prior threshold. Latency has also been optimized via expanded capacity across several Asian regions.
Practical Applications Across Teams
Internal teams at OpenAI are already leveraging the Decisions API across business and product operations. The dots group relies on it to boost interactivity and cut down response times. Meanwhile, the Personal Finance division uses the system to categorize transaction descriptions, and the Communications unit evaluates news coverage themes to determine whether articles reinforce or challenge core narratives.
External users are deploying the tool just as broadly. Use cases range from calculating accurate AI attribution costs to enhancing language-learning engagement. The setup process involves specifying a query, defining permissible outcomes, and supplying context via text, images, or JSON. The endpoint delivers one of three categorical answer types: a probability estimate, a discrete choice with confidence metrics, or a score derived from an ordered rubric.
Real-Time Audio and GPT-Live
With the introduction of native GPT-Live support, the Decisions API can facilitate real-time voice applications. Voice assistants can instantly determine whether to escalate an issue, ask a clarifying question, or redirect a dialogue mid-conversation. Similarly, visual avatars can adapt their expressions in real-time based on live transcript updates, reacting with specific visual cues during natural spoken exchanges.
What it means for developers
For developers building responsive applications, the Decisions API offers a cost-effective and performant way to handle bounded choices using text, images, or JSON inputs. Pricing is set at $0.10 per million input tokens, with cached inputs dropping to $0.01 per million tokens, and zero charges applied to output tokens. Developers can experiment with top AI models cheaply through one API at https://apixoai.online to streamline their infrastructure. Additional options include regional data processing in the US and EU, alongside Zero Data Retention for eligible accounts.
Source: Build more responsive apps with the Decisions API — OpenAI News. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
