OpenAI's Decisions API Goes Live, Taking Aim at TypeSafe's $7.6 Billion Decision-Model Market
OpenAI made its Decisions API publicly available to all customers on Thursday, entering the fast-growing decision-model market dominated by TypeSafe's Jev and cutting response times for some users from over a second to under 200 milliseconds.
OpenAI opened its Decisions API to all customers on Thursday, entering a specialised AI market that TypeSafe's Jev model created and that propelled TypeSafe to a $7.6 billion valuation. Decision models act as ultra-fast "super classifiers", returning a choice from a set of options with a confidence rating in well under a second — far faster than general-purpose chat models. OpenAI built the Decisions API by reconfiguring ChatGPT Luna to prioritise speed and lower inference costs rather than training an entirely new model. Typical response times dropped from one-to-three seconds to 100–180 milliseconds; language-learning app Speak reported cutting its own latency from 1.4 seconds to 0.13 seconds after switching. The API was first announced at OpenAI's developer day on 29 September and now includes voice and visual modalities, with geographically targeted inference to reduce latency globally. On independent benchmarks, Jev retains a strong position — ranking seventh on the AIM System One Index and third on the Artificial Analysis Decision Index, versus eighth and fourteenth for OpenAI's model — and remains the most-queried decision model on OpenRouter. The latest Jev model also now supports images and smart-if statements. OpenAI product lead Nikunj Handa highlighted Luna's multi-modal strengths, noting it is "extremely powerful" for image classification and shows better calibration, while also eyeing further speed gains that could eventually enable per-frame video analysis.
