OpenAI Decisions API vs Jev

OpenAI Decisions API vs Jev is the head-to-head that appeared the moment OpenAI announced its Decisions API at DevDay 2026 — because Jev defined this category first. Both take context plus a finite answer space and return one typed answer with a probability. The difference is what's underneath, how they're calibrated, and — right now — which one you can actually call.

This is a genuine two-horse comparison, and it's worth being fair to both. OpenAI's Decisions API and TypeSafe's Jev solve the same problem — a fast, structured decision your software can act on — but they got there from opposite directions. OpenAI optimized an existing frontier model (GPT-6 Luna) for the decision job; TypeSafe built a model that does nothing but decisions, from scratch. Here's how that plays out.

At a glance

Where OpenAI's Decisions API wins

• It's available in preview from the largest model provider, with the GPT-6 Luna lineage behind it
• Image input — you can send visual context, where Jev takes text state today
• If you're already deep in the OpenAI stack, it's one more endpoint in a platform you know
• ~150ms is fast, roughly 10x quicker than a standard Luna call

Where Jev wins

• It's generally available — you can ship a decision call today, not join a waitlist
• Purpose-built architecture: a typed value in a single non-autoregressive pass, 0% structured-output error by construction
• RLCD calibration — an 0.8 really is right about 80% of the time, so you can trust a threshold like 'escalate if p > 0.8'
• Published, self-serve pricing (sub-cent per decision, output free), and a free browser playground to try first
• Three explicit primitives — choice, score, and a calibrated yes/no (noul) — not just a single selected answer

The real distinction: optimized LLM vs purpose-built model

OpenAI's approach is to take Luna — a general model — and tune it to return a bounded answer fast. That inherits Luna's strengths (including multimodal input) and its lineage. Jev's approach is a model whose only job is the decision: it never learned to generate prose, so there's no token-by-token step to skip and nothing to parse, and it was trained end-to-end with RLCD specifically so its confidence numbers are honest. Neither is automatically better — but if calibration is what you're routing on, a model built and trained for calibrated decisions is a meaningfully different thing from a chat model optimized to be quick.

Which should you use?

Right now the deciding factor is availability: OpenAI's Decisions API is limited preview with no public pricing, while Jev is generally available and self-serve. If you need to ship a decision call this week, build it on Jev — the request shape (context plus a finite answer space) is the same category, so if you later want OpenAI's version for its image input or platform fit, you're porting a well-defined call, not rewriting your logic. We plan to add OpenAI as a provider option here once it's generally available, so the same decide call can target either.

// Build it on Jev today; the shape ports to any decision API later
const res = await fetch("https://jevtypesafeai.com/api/v1/decide", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${process.env.JEV_KEY}`, // jv_live_...
  },
  body: JSON.stringify({
    state: { request: userPrompt },
    questions: {
      route: { type: "choice", instructions: "Route this request.",
               criteria: { fast_model: "cheap + quick", smart_model: "hard reasoning", human: "needs a person" } },
    },
  }),
});
const { answers } = await res.json();
dispatch(answers.route.choice); // one of your three options, with answers.route.confidence

• The New Stack — OpenAI answers TypeSafe's Jev with a Decisions API built on Luna — The framing of OpenAI's Decisions API as a direct response to Jev.
• Vercel — OpenAI Decisions API vs Jev — Side-by-side of the two decision APIs.
• @OpenAIDevs — Decisions API announcement — OpenAI's own announcement post.

FAQ

See also: What is a decision API · OpenAI Decisions API explained · Jev alternatives · What is Jev · Playground