OpenAI has made its Decisions API model broadly available for the first time, stepping into the market for fast, inexpensive intelligence—a space where TypeSafe, a startup founded by a former researcher at the frontier lab, recently reached a $7.6 billion valuation. TypeSafe's Jev model pioneered a category of deep learning models that function as super classifiers: when presented with a set of options, they can rapidly make selections and provide a confidence rating for their choice. Software developers soon discovered that these decision models—sometimes called System One models—could substitute for LLMs across numerous workflows, running faster and at significantly lower cost.

Handa's team raced to release their own version. Instead of constructing a purpose-built decision model, they took ChatGPT Luna and reconfigured it to accelerate inference and reduce cost. "The big unlock with Decisions API has been the speed, shrinking things from one to three seconds to 100, 120, 180 milliseconds," Handa said. "In the future, we may sort of like have specific training to make our models like even better in these ways. But right now, it's just plain old Luna that gets the job done."

EMBED_PLACEHOLDER_0

Unveiled at the company's developer day on September 29, the Decisions API became accessible to all customers on Thursday. OpenAI introduced new voice and visual modalities to the public version and worked to ensure lower latency across global regions by geographically targeting its inference resources. In early trials with customers including Box, Datadog, Ramp, and Speak, engineers used the API to replace OpenAI's LLMs with higher throughput. Andrew Hsu, CTO at Speak, which builds AI language tutors, said the Decisions API enabled the company to reduce the time between actions from 1.4 seconds to .13 seconds. OpenAI also leveraged the product to enhance the responsiveness of its Dots voice assistant.

Still, precisely what constitutes a good decision model remains unclear beyond "it works for you"—optimization across intelligence, cost, calibration, and speed appears to vary by workload. TypeSafe CEO Diogo Almeida is skeptical of benchmarks and encourages people to test the models in their own work, yet even TypeSafe published its own rankings last week alongside the release of its latest Jev model, which adds new features including support for images and smart-if statements. Public benchmarks such as the AIM System One Index and the Artificial Analysis Decision Index tend to favor D1 Ultra, a new decision model 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 ranks eighth and fourteenth, respectively. However, TypeSafe's offerings are currently the most widely queried on OpenRouter.

The most entertaining Jev benchmark, though, is JevMan—a benchmark where decision models play PacMan; the latest Jev model tops that ranking. Felix Wunderlich, chief growth officer at Opper AI, which created the benchmark, said the classic game "is an interesting test because of its nonstop stream of quick, high-stakes decisions. At every junction the model must decide where to go in a few hundred milliseconds." Handa also envisions further speed gains unlocking new use cases: "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."

As these offerings become faster and cheaper, we can expect them to embody the principle of Jev's namesake—an economist who explained how using a resource more efficiently makes it more ubiquitous. Or, as Handa puts it, there will be an expanding "crazy set of use cases" that were hiding in plain sight until the decision models revealed them.