OpenAI Decisions API: What It Is, vs Jev and Laya

What the OpenAI Decisions API does, what OpenAI has and has not published, and how it compares with Jev, Liquid AI d1 and the open-source Laya.

Last updated: Oct 1, 2026

The OpenAI Decisions API is a new OpenAI endpoint that picks one answer from a list you define instead of writing text. You send a question, a fixed set of candidate answers and some context (text or an image), and the OpenAI Decisions API returns the chosen answer. OpenAI announced it at DevDay on September 29, 2026, and runs it on a version of GPT-6 Luna, its low-cost model.

That puts OpenAI in the same category as TypeSafe's Jev, Liquid AI's d1 and the open-source Laya: System One models, which make fast, bounded decisions while a larger model does the slow reasoning. This page collects what OpenAI has published, what it has not, and how the OpenAI Decisions API compares with the System One models you can use today.

OpenAI Decisions API at a glance

What OpenAI has said
AnnouncedDevDay, September 29, 2026
Modela version of GPT-6 Luna
Inputa question, a list of predefined answers, and context as text or images
Outputone answer from your list, not generated text
Speed10x faster than GPT-6 Luna through the regular API: 150 ms against 1.6 s (OpenAI's figures)
Use casesclassify content, route requests, decide an agent's next step
Availabilitylimited preview for selected API customers; broad release announced for the coming days

As of October 1, 2026, OpenAI has not published a price, the request and response format, the maximum number of answers per question, or whether the OpenAI Decisions API returns a probability for each option. The speed figures come from OpenAI and have not been measured independently.

How the OpenAI Decisions API differs from Jev and Laya

OpenAI Decisions APITypeSafe JevLiquid AI d1Laya
Who runs itOpenAI (hosted)TypeSafe (hosted)Liquid AI (hosted)you, or a hosted Laya API
Weightsclosedclosedclosedopen, Apache 2.0
Inputtext or imagestexttexttext, including JSON
Question typespick one answerchoice, score, noulchoice, score, noulchoice, score, noul
Probabilities per optionnot publishedyes, calibratedyes, calibratedyes
API formatnot publishedPOST /v1/systemonePOST /v1/systemone (TypeSafe SDK works)POST /v1/systemone
Pricenot published$0.042 per 1M input tokensfree tier; paid price not publishedfree to self-host
Speed150 ms (OpenAI's figure)410 ms median in our test, network includednot published11 to 54 ms locally in our test
Fine-tuningnot publishednonoyes, on your own labels

The biggest practical differences:

  • Images. The OpenAI Decisions API takes images as context. Jev, d1 and Laya read text only.
  • One wire format, or a new one. Jev, d1 and Laya all speak POST /v1/systemone, so code moves between them by changing the base URL. The OpenAI Decisions API has its own interface, which OpenAI has not documented yet.
  • Probabilities. System One models return a calibrated probability for every option, which is what lets you act automatically above a threshold and send the rest to a human. OpenAI has only described the OpenAI Decisions API as returning a selection.
  • Control. Laya's weights are open: you can run it on your own servers, keep data in-house and fine-tune it. The three hosted APIs, OpenAI included, cannot be self-hosted.

Which one should you use?

Your situationStart with
Decisions over screenshots, photos or other imagesOpenAI Decisions API, once you have access
You already use OpenAI and want one bill and one accountOpenAI Decisions API
You need probabilities to set thresholds and route uncertain casesJev, d1 or Laya
You want a documented API today, with SDKs and per-token pricingJev
Data must stay on your servers, or you want to fine-tuneLaya, self-hosted
You want open weights without running serversa hosted Laya API

If you are evaluating now, build against the /v1/systemone format with Jev or Laya and keep your questions in one place. When OpenAI publishes the OpenAI Decisions API format, you can add it as one more backend and compare all of them on the same labelled sample. Our System One benchmark shows how much accuracy differs between models on identical questions.

Is the OpenAI Decisions API a Jev competitor?

Yes, for hosted use. Both target the same job: a fast, cheap decision inside an agent or a pipeline, so the expensive model is only called when it is needed. OpenAI has not said the launch is a response to Jev. For teams already on OpenAI, the Decisions API removes the need for a second provider; for teams that need calibrated probabilities, an open model or a documented API today, Jev and Laya still cover needs that OpenAI has not addressed yet.

FAQ

Is the OpenAI Decisions API available now?

Only as a limited preview for selected API customers. OpenAI announced broad availability for the coming days after DevDay on September 29, 2026.

How much does the OpenAI Decisions API cost?

OpenAI has not published a price. Do not assume regular GPT-6 Luna token rates: it is a separate endpoint and OpenAI said it will share details at broad release. For reference, Jev costs $0.042 per 1M input tokens and Laya is free to self-host.

Does the OpenAI Decisions API return probabilities?

OpenAI has described the output as the selected answer. Whether it also returns a probability for each option is not published. Jev, d1 and Laya return a probability for every option.

Can I use the TypeSafe SDK or Jev code with the OpenAI Decisions API?

Not as far as OpenAI has shown. Jev, Liquid AI d1 and Laya share the /v1/systemone format; the OpenAI Decisions API has its own interface, not yet documented.

Is there an open-source alternative to the OpenAI Decisions API?

Yes. Laya is an open-weight System One model under Apache 2.0, with a 100+ language checkpoint, and other open models are listed on System One models compared and Jev alternatives.

Sources

Last verified: October 1, 2026. We will update this page when OpenAI publishes documentation and pricing.