OpenAI opens fast AI Decisions API to all customers

OpenAI on Thursday, Oct. 22, made its Decisions API generally available. The tool is built on GPT-6 Luna and answers bounded questions "up to 10x faster than GPT‑6 Luna through the Responses API," according to the company. The release puts OpenAI directly in competition with TypeSafe, a startup whose Jev model created the category of fast, cheap "decision" models and helped make the company worth $7.6 billion, TechCrunch reported.openai.com+1

What changed since beta

The API skips full-text generation. Developers write a question, list the allowed answers and supply context as text, images or JSON. The API then returns a probability, a choice or a rubric score that an application can act on, OpenAI said. OpenAI's changelog shows the API entered beta on Oct. 6. Since then, OpenAI has added a native GPT-Live voice integration, prompt caching, and inference residency in the US and EU. It also doubled the limit to 400 questions per request, the company said. Pricing is $0.10 per million input tokens and $0.01 per million cached tokens, and there is no charge for output tokens.community.openai.com+2

Several early customers reported speed gains. Box ↗ 0.07% said route selection was about 5x faster. Datadog ↗ 3.93% said it saw 40x faster p95 latency than frontier models when scoring the risk of agent-generated commands. Language-learning app Speak said its time to next action fell from 1.4 seconds to 0.13 seconds.openai.com

The race with Jev

Nikunj Handa, OpenAI's API product lead, told TechCrunch that colleagues who saw Jev reacted with "wow, we we have like a million use cases for this." He said OpenAI did not train a dedicated decision model. Instead, it reconfigured Luna to run faster and cheaper. "Right now, it's just plain old Luna that gets the job done," he said.techcrunch.com

OpenAI does not lead the field yet. TechCrunch reported that the AIM System One Index and the Artificial Analysis Decision Index tend to favor Liquid's D1 Ultra and Base Lab's Elephant-1. Jev ranks seventh and third on those indexes, while OpenAI's model ranks eighth and fourteenth. TypeSafe's models are also the most-queried on OpenRouter. Jev is also cheaper, at a listed $0.042 per million input tokens, but it accepts only text, one developer analysis noted.cellcog.ai+1

Betting on images and speed

Handa said image handling could give OpenAI an edge. "Luna is extremely powerful and has been showing better calibrations" for image classification, he said. He also suggested that lower latency could lead to new uses. "Can we get this to something like you know 60 milliseconds, and then you could imagine like an API like call like for every single frame in like a video?".techcrunch.com