OpenAI Launches General Availability of Decisions API
Built with GPT-6 Luna, the Decisions API helps applications and agents process text, images, and structured inputs up to 10 times faster at a lower cost.
Radwan AbdulhadiEditor and technology writer
General Availability Announcement
OpenAI has officially announced that the Decisions API is now generally available for developers and enterprises building applications that require fast, focused, and low-cost decisions. Applications and automated agents frequently need to make numerous small choices throughout workflows, such as classifying input data, routing requests, or deciding the next operational step. While individual choices are often simple, their cumulative latency and cost can scale rapidly across heavy workloads.
Built utilizing GPT‑6 Luna, the Decisions API focuses model intelligence specifically on questions featuring bounded answers. This enables applications to achieve near real-time decision-making from text, images, and structured inputs up to 10 times faster than processing through standard endpoints.
New Capabilities and Enhancements
Since its initial public beta phase, OpenAI has introduced several key updates to the platform. These enhancements include a native integration, prompt caching options for reused context, and inference data residency choices in both the United States and the European Union. Developers can now submit native JSON inputs, incorporate images directly into question instructions, and evaluate up to 400 questions per request, doubling the previous limit.
Additional infrastructure investments have further optimized latency, including newly added processing capacity across several regions in Asia. For developers looking to experiment or check exact cost breakdowns, OpenAI provides a dedicated pricing and availability guide outlining current rates and caching details.
Supported Answer Types and Core Use Cases
Using the Decisions API requires developers to define a specific question alongside its allowed answers, followed by context supplied as text, images, or native JSON data. The platform returns one of three distinct answer types to dictate application logic: probabilities, discrete choices, or ordered scores.
The supported output categories include probability values ranging from 0 to 1 to estimate whether a condition holds true, categorical choices selected from supplied custom values, and rubric scores evaluated against ordered levels. Internal OpenAI teams and external customers are already applying these mechanisms to various use cases, such as categorizing transaction descriptions, reducing response latency, classifying communication coverages, and improving interactive educational software.
Real-Time Voice Integration with GPT-Live
The release also bridges fast decision-making with conversational audio experiences. Through integrations with GPT‑Live, developers can build natural voice applications featuring smooth interruption handling that keeps dynamic conversations flowing smoothly.
The Decisions API integrates natively with GPT‑Live, enabling voice agents to make rapid, focused decisions during live interactive sessions. Customer support agents can instantly decide whether to route a call, issue a clarifying question, or escalate an issue. Similarly, virtual avatars can dynamically alter facial expressions based on user inputs to deliver more engaging, responsive virtual environments.
Pricing Structure and Developer Resources
The Decisions API pricing model is structured around input tokens without any associated output-token charges. Base input processing costs $0.10 per million tokens, while cached input tokens drop to $0.01 per million tokens. Regional processing capabilities are deployed in the US and EU, with Zero Data Retention options available for eligible commercial customers.
Developers can begin testing queries, texts, and images directly within the Decisions Playground to evaluate recurring choices before writing production code.
Sources
- OpenAI NewsBuild more responsive apps with the Decisions API
Continue chronologically
Related entity coverage
Related stories
OpenAI Expands GPT-6 Series with Sol and Luna Models
OpenAI expands its GPT-6 universe by introducing Sol and Luna, two new cost-efficient models offering high performance across coding and professional tasks.
OpenAI Releases Practical Guide for GPT-6 Integration
OpenAI details model selection, reasoning effort tuning, context optimization, and production best practices for the GPT-6 model suite.
OpenAI Rolls Out GPT-6 to All Tiers of ChatGPT
OpenAI is rolling out GPT-6 to all ChatGPT tiers, bringing a new Intelligent UI capability that delivers interactive visual responses and faster answers.
Qodo Brings Organizational Context and Code Review to Codex
Discover how Qodo's new Codex plugin integrates organizational knowledge, coding standards, and independent review to catch P0 bugs early.
