# OpenAI Makes Its Decisions API Generally Available, Cutting Latency to Roughly 150 Milliseconds

*2026-10-25 03:01 UTC · AI · 1 min read*

> OpenAI opened its fast Decisions API to all customers, built by reconfiguring ChatGPT Luna rather than training a new model.

OpenAI made its Decisions API available to all customers on Thursday, after announcing it at its September 29 developer day. The model was built by reconfiguring ChatGPT Luna for faster, cheaper inference rather than training a dedicated decision model. Latency fell from about 1-3 seconds to roughly 100-180 milliseconds, according to product lead Nikunj Handa.

The public version adds voice and visual modalities and uses regionally targeted inference resources. Early customers include Box, Datadog, Ramp and Speak; Speak CTO Andrew Hsu said the time between actions dropped from 1.4 seconds to 0.13 seconds. OpenAI wants to reach about 60 milliseconds, which would allow per-frame video calls. Pricing was not disclosed.

The launch targets TypeSafe, whose Jev model popularized the category of fast "super classifier" decision models and is the most widely queried decision model on OpenRouter. Rankings are mixed: Jev placed 7th and OpenAI 8th on the AIM System One Index, and 3rd versus 14th on the Artificial Analysis Decision Index.

## Sources

- [techcrunch.com](https://techcrunch.com/2026/10/22/openai-makes-its-jev-competitor-widely-available-as-decision-model-competition-heats-up/)

---
TechTools News · https://techtools.cz/tools/news/article/2574-openai-makes-its-decisions-api-generally-available-cutting-latency-to-roughly
Česká verze: https://techtools.cz/tools/news/clanek/2648-openai-zpristupnilo-vsem-decisions-api-odezva-klesla-zhruba-na-150-milisekund