OpenAI released its Decisions API to every user on 2026‑10‑22. The model is a re‑tuned version of ChatGPT Luna that runs in 100‑180 ms instead of a few seconds. OpenAI added voice and visual support and spread inference servers worldwide to keep latency low. Early adopters like Box, Datadog, Ramp and Speak say the API lets them replace standard LLM calls with much faster decisions. The move puts OpenAI into the growing “decision model” market that TypeSafe’s Jev model helped create.
Why it matters
Developers can replace slower LLM calls with the fast, cheap Decisions API, cutting response times in real‑world apps such as Speak from 1.4 s to 0.13 s.
