Saturday, Oct 24, 2026 About UsContact UsPrivacy PolicyAdvertise
Subscribe
AI News

OpenAI Decisions API Now Widely Available: Fast, Cheap AI Decisions Rival Jev

OpenAI's Decisions API, a fast and low-cost decision model built on ChatGPT Luna, is now available to all customers as competition with TypeSafe's Jev heats up.

OpenAI Decisions API graphic showing fast AI choosing between options with a stopwatch
Graphic: New New Technology
Quick answerThe OpenAI Decisions API is a fast, low-cost AI service that picks between a set of options and returns a confidence score. Built on a speed-tuned ChatGPT Luna, it cuts response times from 1–3 seconds to about 100–180 milliseconds and became available to all customers on October 22, 2026.

The OpenAI Decisions API is now available to every OpenAI customer. First announced at OpenAI's developer day on September 29, 2026, the service opened widely on October 22. It puts OpenAI squarely into a new and fast-growing market: decision models that make quick, cheap choices instead of writing long answers.

What Is the OpenAI Decisions API?

Think of a decision model as a super classifier. You give it a set of options. It picks one in a fraction of a second and tells you how confident it is. Developers have found these "System One" models can replace large language models in many workflows, at much lower cost and higher speed.

The category took off with Jev, a model from TypeSafe, a startup founded by a former OpenAI researcher. Jev's success helped make TypeSafe a $7.6 billion company.

How OpenAI Built It So Fast

OpenAI did not train a brand-new model. Its API team took ChatGPT Luna and reconfigured it for faster inference and lower cost. Nikunj Handa, product lead for OpenAI's APIs, said the big unlock was speed, shrinking response times from one to three seconds down to about 100 to 180 milliseconds.

Latency comparison for the OpenAI Decisions API: 1-3 seconds for standard LLM calls versus about 100-180 milliseconds
Decisions API speed gains (Graphic: New New Technology)

The public version adds voice and visual inputs. OpenAI also placed inference resources in regions around the world to lower latency for global users.

Real-World Results From Early Customers

  • Speak, an AI language tutor, cut the time between actions from 1.4 seconds to 0.13 seconds.
  • Box, Datadog and Ramp used it to replace slower LLM calls and boost throughput.
  • OpenAI's own Dots voice assistant became more responsive using the API.

OpenAI Decisions API vs Jev and Other Rivals

Competition is intense. On public leaderboards, OpenAI is not yet on top.

ModelMakerNotable strength
D1 UltraLiquidLeads several public decision benchmarks
Elephant-1Base LabStrong benchmark scores
Jev (latest)TypeSafeMost queried on OpenRouter; top of the JevMan Pac-Man test
Decision ModelOpenAIMultimodal power, especially image classification

On the AIM System One Index and Artificial Analysis Decision Index, Jev ranks seventh and third, while OpenAI's model ranks eighth and fourteenth. TypeSafe's CEO has urged developers to trust real-world trials over benchmarks.

Why It Matters

Many AI products feel slow because every step waits on a large language model. Decision models remove that delay for simple choices. Faster and cheaper calls mean AI can be used in places where it was too costly before, such as real-time voice apps, live moderation and high-volume support systems. For developers, the OpenAI Decisions API adds a strong option to a market that was dominated by startups.

Best Use Cases for Decision Models

  1. Routing support tickets to the right team.
  2. Content moderation and spam filtering.
  3. Real-time voice assistant turn-taking.
  4. Image classification in product catalogues.
  5. Game AI and other fast, repeated choices.

What Comes Next

Handa hopes to push speed to around 60 milliseconds, fast enough to run a call on every frame of a video. As decision models get cheaper, expect them to appear everywhere, much like the economist Jevons predicted when efficiency drives wider use. For another fresh model launch, see Canva Craft, Canva's first image model.

Explore more launches in our AI Model Releases section.

Key takeaways

  • OpenAI's Decisions API became generally available on October 22, 2026, with new voice and visual modalities.
  • It cuts response times from 1–3 seconds to roughly 100–180 milliseconds using a reconfigured ChatGPT Luna.
  • TypeSafe's Jev still leads on popularity, while Liquid's D1 Ultra tops several public benchmarks.

Frequently asked questions

What is a decision model?

A decision model, sometimes called a System One model, acts like a super classifier. Given a set of options, it quickly chooses one and gives a confidence rating.

How fast is the OpenAI Decisions API?

OpenAI says it reduces latency from one to three seconds down to around 100–180 milliseconds.

Which companies have tested it?

Early customers include Box, Datadog, Ramp and Speak. Speak cut time between actions from 1.4 seconds to 0.13 seconds.

Sources

Facts last verified Oct 23, 2026.

Tags: OpenAIDecisions APIAI ModelsDevelopersJev

Related articles