Decision API
An API introduced by OpenAI that focuses model intelligence on problems with a finite, user-defined set of possible answers. It supports real-time decision-making and is applicable to use cases such as content classification, request routing, and agent action selection.
Timeline (last 90 days)
Because RAG-type applications have a larger number of input tokens, reducing the unit price of input tokens has a more pronounced cost optimization effect on RAG-type applications than on ordinary question-answering applications.
The Decision API has a built-in retrieval-augmented generation (RAG) pipeline that first fetches the latest information from the internet or specified data sources, then combines language model reasoning to output decision-oriented conclusions directly usable by programs.
Input token billing for the Decision API typically covers the user's incoming prompt and retrieved recall content, while output tokens correspond to the final result returned by the model.
What was officially announced this time is only the input token price; output-side pricing, rate limits, and other billing details were not explicitly mentioned in the announcement.
Perplexity has reduced the input token price of the Decision API to 2 cents per million tokens.
This Decision API price reduction cuts the previous price in half again (a 50% reduction).
This price adjustment is a purely pricing change rather than a feature-level upgrade.
The Decision API is a dedicated interface launched by Perplexity whose core capability is to deliver structured judgments or recommendations on user questions based on real-time online retrieval, rather than simply returning raw search results.
OpenAI launched the Decision API real-time classifier, which the company states can make decisions up to 10 times faster than GPT-6 Luna via the Response API.
OpenAI released the Decision API, a general-purpose classifier API with image capabilities, which will be available approximately one week after the announcement.
1 more timeline events
All Facts (11)
Because RAG-type applications have a larger number of input tokens, reducing the unit price of input tokens has a more pronounced cost optimization effect on RAG-type applications than on ordinary question-answering applications.
50%UnverifiedThe Decision API is a dedicated interface launched by Perplexity whose core capability is to deliver structured judgments or recommendations on user questions based on real-time online retrieval, rather than simply returning raw search results.
50%UnverifiedThe Decision API has a built-in retrieval-augmented generation (RAG) pipeline that first fetches the latest information from the internet or specified data sources, then combines language model reasoning to output decision-oriented conclusions directly usable by programs.
50%UnverifiedInput token billing for the Decision API typically covers the user's incoming prompt and retrieved recall content, while output tokens correspond to the final result returned by the model.
50%UnverifiedWhat was officially announced this time is only the input token price; output-side pricing, rate limits, and other billing details were not explicitly mentioned in the announcement.
50%UnverifiedThis Decision API price reduction cuts the previous price in half again (a 50% reduction).
50%UnverifiedThis price adjustment is a purely pricing change rather than a feature-level upgrade.
50%UnverifiedDecision API focuses model intelligence on a set of user-defined questions with limited answers, where developers provide context via text or images and the system returns answers that can be used for content classification, request routing, or selecting an agent's next action; it is currently in limited preview.
50%UnverifiedPerplexity has reduced the input token price of the Decision API to 2 cents per million tokens.
50%UnverifiedOpenAI launched the Decision API real-time classifier, which the company states can make decisions up to 10 times faster than GPT-6 Luna via the Response API.
50%UnverifiedOpenAI released the Decision API, a general-purpose classifier API with image capabilities, which will be available approximately one week after the announcement.
50%