OpenAI Opens Decisions API Widely as Rivals Challenge Jev

OpenAI opened its Decisions API to all customers on Thursday, moving a company built on large language models into a category that rival startup TypeSafe created with its Jev system, according to TechCrunch. The product, OpenAI said, lets developers swap slower, more expensive language models for a faster classifier-style system in workflows that need a quick choice rather than a generated answer.

The launch follows a developer-day preview on September 29 and arrives as TypeSafe, founded by a former OpenAI researcher, has grown into a company valued at $7.6 billion on the strength of Jev, TechCrunch reported. Independent benchmark trackers now rank several entrants, and OpenAI's own figures on speed gains have not yet been independently verified outside the customers it named.

What is OpenAI's Decisions API and why does it matter?

Jev and similar "System One" models are built to take a defined set of options and quickly return a choice along with a confidence score, rather than generating open-ended text the way a large language model does. Developers told TechCrunch they found these models could replace LLMs in many workflows, running faster and at lower cost.

Nikunj Handa, product lead for OpenAI's APIs, said colleagues who saw Jev concluded the company had "like a million use cases for this," according to TechCrunch. Rather than train a new model from scratch, Handa said his team reconfigured ChatGPT Luna to speed up inference and cut cost. "The big unlock with Decisions API has been the speed, shrinking things from one to three seconds to 100, 120, 180 milliseconds," Handa told TechCrunch. "Right now, it's just plain old Luna that gets the job done."

How does OpenAI's model compare to TypeSafe's Jev?

TypeSafe released an updated version of Jev last week with support for images and what the company calls smart-if statements, along with its own internal rankings, TechCrunch reported. TypeSafe CEO Diogo Almeida said he distrusts public benchmarks and has urged developers to test models directly on their own workloads rather than rely on leaderboards.

OpenAI's new release adds voice and visual modalities and works to route inference to servers closer to users to cut latency across regions, according to the company's account to TechCrunch.

What do early customers say about speed gains?

Box, Datadog, Ramp and Speak tested the Decisions API before Thursday's wider launch, TechCrunch reported. Andrew Hsu, chief technology officer at Speak, which builds AI language tutors, told TechCrunch the API cut the gap between user actions from 1.4 seconds to 0.13 seconds. OpenAI also used the system internally to speed responses in its Dots voice assistant, the company said.

How do independent benchmarks rank the new decision models?

Two public trackers, the AIM System One Index and the Artificial Analysis Decision Index, currently favor two newer entrants: D1 Ultra, released last week by Liquid, and Elephant-1, from Base Lab, TechCrunch reported. The latest Jev model still ranks seventh and third, respectively, on those indices. TechCrunch's report did not include a ranking position for OpenAI's Decision Model on either index.

What remains uncertain about measuring decision models?

Even the companies building these systems caution that no single standard defines a good decision model. TechCrunch reported that performance trade-offs among intelligence, cost, calibration and speed vary by workload, which is why Almeida recommends direct testing over benchmark comparisons. For now, the competing claims — OpenAI's internal latency figures, TypeSafe's own rankings, and the two independent indices — have not been reconciled into one comparison, and TechCrunch's reporting does not resolve which system performs best across the board.

Reported figures

• OpenAI says its Decisions API cut response times from one-to-three seconds down to 100, 120 and 180 milliseconds, depending on task, per Handa.
• Speak's CTO said a customer-facing delay fell from 1.4 seconds to 0.13 seconds after adopting the API.
• TypeSafe, maker of Jev, is valued at $7.6 billion, according to TechCrunch.
• The latest Jev model ranks seventh on the AIM System One Index and third on the Artificial Analysis Decision Index.

Timeline

• Sept. 29, 2026: OpenAI previews the Decisions API at its developer day.
• Week of Oct. 15, 2026: TypeSafe releases an updated Jev model with image support and smart-if statements, plus its own rankings.
• Oct. 22, 2026 (Thursday): OpenAI's Decisions API becomes available to all customers, with added voice and visual modalities.

Artiglio is A coming-soon iPhone chief of staff for briefings and drafts. Not on the App Store yet.