OpenAI makes Decisions API widely available

OpenAI has made its Decisions API model widely available, entering a market shaped by TypeSafe’s Jev, a decision model that quickly selects from options and provides a confidence rating. TypeSafe, founded by a former researcher at a frontier lab, was valued at $7.6 billion, according to the source.

OpenAI product lead Nikunj Handa said the company adapted ChatGPT Luna to speed up inference and reduce costs rather than build a specially calibrated decision model. He said the change cut response times from one to three seconds to 100, 120 or 180 milliseconds. The API was announced at OpenAI’s developer day on September 29 and became available to all customers on Thursday. The public version also adds voice and visual capabilities.

Customers including Box, Datadog, Ramp and Speak have tested the API. Speak CTO Andrew Hsu said it reduced the time between actions at the language-learning company from 1.4 seconds to 0.13 seconds. OpenAI also used the product to make its Dots voice assistant more responsive.

Public rankings give mixed results. The latest Jev model ranks seventh on the AIM System One Index and third on the Artificial Analysis Decision Index; OpenAI’s Decision Model ranks eighth and fourteenth, respectively. The source says the rankings tend to favor Liquid’s D1 Ultra and Base Lab’s Elephant-1. TypeSafe’s models are the most queried on OpenRouter, while Jev leads the JevMan PacMan benchmark.

Handa said image classification could be a strength for Luna and suggested that faster response times, potentially 60 milliseconds, could enable new uses such as processing every video frame. The source notes that what makes a decision model effective can vary by workload.