OpenAI Releases Decisions API With Enhanced Features
OpenAI's Decisions API is now generally available, allowing applications to classify text and images with increased efficiency. The API's new features include a doubled limit of 400 questions per request and prompt caching.
Original reporting · searchenginewatch.com
Original: “OpenAI makes Decisions API generally available with 400-question requests”
Read “OpenAI makes Decisions API generally available with 400-question requests” at Search Engine Watch →
How this was written. The 5:45 Ops Desk wrote this article with a language model from Search Engine Watch’s reporting (model: @cf/meta/llama-3.3-70b-instruct-fp8-fast); it describes that reporting rather than adding to it. The original, linked above, is the full story.
Key points from Search Engine Watch’s report
- OpenAI's Decisions API is now generally available, with a doubled limit of 400 questions per request.
- The API uses GPT-6 Luna to classify text and images, route customer requests, or select an agent's next action from bounded answers.
- The API's pricing model is based on the number of tokens processed, with discounts for cached input.
- Prompt caching can significantly reduce processing costs by reusing unchanged portions of the shared input.
- The API's flexibility and scalability make it an attractive option for developers and businesses
OpenAI has made its Decisions API generally available, marking a significant milestone for the company. According to Search Engine Watch, the API uses GPT-6 Luna to classify text and images, route customer requests, or select an agent's next action from bounded answers. This development is expected to have a major impact on the industry, as it enables applications to make more informed decisions.
The Decisions API allows applications to choose the form of the answer, with three distinct answer types: predicate, choice, and scores. A predicate estimates whether a condition is true, such as whether a product photo shows damage. This feature is particularly useful for applications that require a simple yes or no answer.
The choice answer type selects a category, such as billing or shipping, while scores apply ordered levels, as in the example of grading damage against ordered severity levels. The API's documentation provides clear examples of how these answer types can be used in real-world applications. Search Engine Watch reported that the API's flexibility is one of its key strengths.
One of the key features of the Decisions API is its ability to handle independent questions that share one input. However, questions that depend on earlier answers require separate requests. This means that a 400-question call evaluates shared evidence rather than running a sequence of dependent tasks. The API's design allows for efficient processing of large amounts of data.
The October 22 release of the Decisions API also adds prompt caching, native JSON input, and inference residency in the US and EU. Prompt caching reuses an unchanged portion of the shared input, such as a reference document, which can significantly reduce processing costs. The reusable prefix must contain at least 1,024 tokens, and question text does not count toward that threshold.
OpenAI's pricing model for the Decisions API is based on the number of tokens processed. Uncached input is charged at $0.10 per million tokens, while cached input is charged at $0.01 per million. Cache writes have separate charges, and regional processing premiums can apply. However, output tokens are not charged, which can help reduce costs for applications that require large amounts of output data.
The general availability of the Decisions API is expected to have a significant impact on the industry, as it enables applications to make more informed decisions. With its enhanced features and flexible pricing model, the API is likely to attract a wide range of developers and businesses. As the API continues to evolve, it will be interesting to see how it is used in real-world applications.
Questions and answers
What is the Decisions API and how does it work?
The Decisions API uses GPT-6 Luna to classify text and images, route customer requests, or select an agent's next action from bounded answers. It allows applications to choose the form of the answer, with three distinct answer types: predicate, choice, and scores.
How does prompt caching work in the Decisions API?
Prompt caching reuses an unchanged portion of the shared input, such as a reference document, which can significantly reduce processing costs. The reusable prefix must contain at least 1,024 tokens, and question text does not count toward that threshold.
How is the Decisions API priced?
The API's pricing model is based on the number of tokens processed, with uncached input charged at $0.10 per million tokens and cached input charged at $0.01 per million. Cache writes have separate charges, and regional processing premiums can apply.
