OpenAI’s GPT-6 Luna-Powered Decisions API Is Now Generally Available for Developers With Native GPT-Liv...
OpenAI's Decisions API was available in public beta for some time. Now, the technology company led by Sam Altman has announced that the Decisions API is generally available. This special tool lets developers build AI agents and apps that can answer questions that come up often and make small decisions. Developers can give it images, text, and information in a structured format called JSON. The Decisions API uses OpenAI's GPT-6 Luna AI model. OpenAI says it now also works directly with GPT-Live. Since the public beta launch, the company has added prompt caching in certain markets. This feature lets the tool reuse information it has already received.
Decisions API Price and Features
In a blog post on Thursday, OpenAI announced that developers in certain markets around the world can now use the Decisions API through its general release. They can access it in the Decisions Playground. The tool runs on GPT-6 Luna and costs $0.10 (about Rs. 10) for every million input tokens. Tokens are small pieces of text or information that AI systems process. When developers reuse saved information through caching, the cost drops to $0.01 (about Rs. 1) per million tokens. The Decisions API is also available to eligible customers who use Zero Data Retention, a setting that prevents their data from being stored.
Developers can use the GPT-6 Luna-powered Decisions API to help their AI agents and apps make small decisions that happen again and again. For example, it can put information into categories, organise information, send requests to the right place, and choose what an app should do next. OpenAI says the Decisions API can make decisions almost instantly using images, text, and structured information. The company claims it can do this up to 10 times faster than using the GPT-6 Luna AI model through the Responses API.
OpenAI also says developers can send information directly in JSON format and include images with their questions and instructions. They can now ask up to 400 questions in a single request to the Decisions API. When the tool first launched in public beta, the limit was 200 questions per request. Besides raising this limit, OpenAI has added direct GPT-Live integration, prompt caching to reuse information, and inference residency in the US and EU. Inference residency means developers can choose to have the AI process their requests in those regions.
To use the Decisions API, a developer gives it questions, a list of possible answers, and background information. The API then uses that background information to choose the answer that best fits each question. It sends the chosen answer back to the app or AI agent so it can take the next step. With the new GPT-Live integration, the Decisions API can also help apps, voice-based AI agents, and AI customer support systems decide what to do during conversations. For example, it can decide whether to ask someone for more details, send their request somewhere else, or pass a problem to someone who can provide more help.