OpenAI Makes Its Rival to Jev More Widely Available as Decision-Model Competition Intensifies

OpenAI has made its Decisions API model available to all customers for the first time, putting ChatGPT Luna to work as a fast, low-cost classifier. The move places OpenAI directly in a market that has grown around rapid decision-making models, a category that TypeSafe helped define with its Jev model and that has turned the startup, founded by a former researcher at the frontier lab, into a $7.6 billion company.

Background: Super Classifiers Create a New AI Layer

TypeSafe’s Jev model established a category of deep learning models that function as super classifiers. Given a set of options, these models quickly choose one and provide a confidence rating for the decision. That design proved useful for software developers, who found that decision models, sometimes described as System One models, could replace large language models (LLMs) in many workflows while running faster and at dramatically lower cost.

Nikunj Handa, the product lead for OpenAI’s APIs, told TechCrunch that colleagues saw Jev and immediately recognized the opportunity, imagining “a million use cases” for the technology. His team sprinted to create an OpenAI version. Instead of building a specifically calibrated decision model from scratch, it took ChatGPT Luna and reconfigured it to speed up inference and lower cost.

Handa described speed as the central change. He said response times shrank from one to three seconds to 100, 120, or 180 milliseconds. He added that future training may improve the model for these tasks, but the current product is “plain old Luna” adapted for fast decisions.

Development: From Developer Day to Broad Availability

OpenAI announced the Decisions API at its developer day on September 29, and it became available to all customers on Thursday. The public version added new voice and visual modalities, extending the product beyond text-only classification. OpenAI also worked to lower latency in regions around the world by targeting its inference resources geographically.

In initial trials with customers including Box, Datadog, Ramp, and Speak, engineers used the API to replace OpenAI’s LLMs with higher throughput. Andrew Hsu, the CTO at Speak, which builds AI language tutors, said the Decisions API allowed the company to cut the time between actions from 1.4 seconds to .13 seconds. OpenAI also used the product to improve the responsiveness of the Dots voice assistant.

Still, there is no settled definition of what makes a good decision model other than whether it works for a given workload. Optimization across intelligence, cost, calibration, and speed appears to vary by use case. TypeSafe CEO Diogo Almeida distrusts benchmarks and urges developers to trial models in their own work. Even so, TypeSafe released its own rankings last week alongside its latest Jev model, which includes support for images and smart-if statements.

Latest Progress: Benchmarks, Competition, and Wider Use

Public benchmarks such as the AIM System One Index and the Artificial Analysis Decision Index tend to favor D1 Ultra, a new decision model that Liquid released last week, and Base Lab’s Elephant-1. The latest Jev model remains competitive on both indices, placing seventh and third, while OpenAI’s Decision Model falls eighth and fourteenth, respectively. For now, TypeSafe’s offerings are the most widely-queried on OpenRouter.

A more playful test is JevMan, a benchmark in which decision models play Pac-Man. The latest Jev model tops that ranking. Felix Wunderlich, chief growth officer at Opper AI, which built the benchmark, said the classic game is interesting because it presents a nonstop stream of quick, high-stakes decisions. At every junction, the model must choose where to go in a few hundred milliseconds.

OpenAI’s advantage may lie in multi-modality. For image classification, Handa said Luna is extremely powerful and has been showing better calibrations. He also noted that OpenAI customers want the Decisions API to become more flexible about using images as outputs as well as inputs.

Handa sees further speed gains opening new possibilities. If the API reaches around 60 milliseconds, he suggested, developers could imagine making a call for every single frame in a video. As these offerings become faster and cheaper, they may live up to the idea behind Jev’s namesake, an economist who explained how using a resource more efficiently makes it more ubiquitous. Handa’s broader expectation is that decision models will reveal an expanding “crazy set of use cases” that were hiding in plain sight.

The current stage is OpenAI’s shift from a targeted decision model concept to broadly available infrastructure: a faster, cheaper classification layer that now faces benchmark competition, multimodal expectations, and new real-time use cases.

Meta Description: OpenAI's Decisions API is now available to all customers, using ChatGPT Luna for fast, low-cost classification workloads across global regions.

Tags: OpenAI, Decisions API, ChatGPT Luna, TypeSafe Jev, decision models

Featured image: Abstract technology style, glowing decision paths and speed gauges; no real people, no company logos, no text.