OpenAI makes its Jev competitor widely available as decision model competition heats up

OpenAI has made its Decisions API generally available, offering a low-latency, lower-cost model for tasks that require choosing among options and reporting confidence. First announced at OpenAI’s developer day on September 29, the API is based on a reconfigured version of ChatGPT Luna rather than a purpose-trained decision model. The public release adds voice and visual capabilities and geographically targeted inference resources intended to reduce latency worldwide. Decision models, also called System One models, have gained traction as faster alternatives to large language models for classification and routing workflows. OpenAI product lead Nikunj Handa said Decisions API can reduce response times from one to three seconds to roughly 100 to 180 milliseconds. Speak CTO Andrew Hsu said the language-tutoring company reduced time between actions from 1.4 seconds to 0.13 seconds, while Box, Datadog, Ramp, and Speak tested the API to increase throughput. OpenAI also used it to improve responsiveness in its Dots voice assistant. Competition has intensified after TypeSafe, founded by a former OpenAI researcher, reached a $7.6 billion valuation around its Jev decision model. Public indices rank Liquid’s D1 Ultra and Base Lab’s Elephant-1 above OpenAI’s model, while Jev remains among the leading systems and is the most queried decision-model offering on OpenRouter. OpenAI says Luna may hold an advantage in image classification and calibration.