Microsoft-Decision-1 vs Jev

Microsoft-Decision-1 is the decision-scoring model Satya Nadella unveiled on 9 October 2026 — Microsoft's entry into the category Jev defined. Like Jev, it takes context plus a fixed answer space and returns a structured result with a probability per option — yes/no, a pick from a set, or a score — instead of generated text. It's the biggest vendor yet to ship a decision model, which makes this comparison worth doing honestly: not whether the category is real (Microsoft just validated it), but which model to reach for today.

If you're weighing Microsoft-Decision-1 against Jev, both answer the same question in the same shape: a class, a yes/no, or a rating, each with a calibrated probability, in a single pass your software can act on. The real choice is about how each was built and where you can run it — not about whether either can decide.

What Microsoft-Decision-1 is

Microsoft describes it as a decision-scoring model for routing, classification, prioritization, verification and workflow control. Rather than writing prose, it emits structured output with a probability score for each option — the three task shapes reported are yes/no, multiple-choice, and scoring. It was further-trained from Alibaba's open-source Qwen3.5-9B, and Microsoft says future versions will be based on its own MAI model and on OpenAI models. It's available first through Microsoft Foundry, with OpenRouter coming soon.

Specs, side by side

The Microsoft-Decision-1 column uses Microsoft's own announcement figures; the Jev column uses TypeSafe's. Every performance number here is vendor-reported, not an independent benchmark.

Microsoft's speed and cost claims — read them honestly

Microsoft's headline numbers come from internal teams, not a third-party lab. Its Xbox research team reported classifying more than 10,000 pieces of gamer feedback (from surveys, Steam and X) at roughly a two-hundredth of the cost and more than 14x the speed of GPT-6 Sol, with similar quality; a Copilot team reported about 100x the speed of GPT-5.6 Luna at similar response-quality. Those are impressive but vendor-measured against specific large models on Microsoft's own tasks — the same caveat that applies to TypeSafe's own 40–200x figure for Jev. The only benchmark that settles it is your decision, on your data.

The real distinction: post-trained open model vs purpose-built

Microsoft's approach takes an existing open model — Qwen3.5-9B — and post-trains it to return bounded, scored answers fast. Jev's approach is a model whose only job is the decision: it never learned to generate prose, so there is 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 — an 0.8 is right about 80% of the time. Neither is automatically better, but if you route on the probability itself, a model built and calibrated for that is a different thing from a general model tuned to be quick.

When to pick which

• Reach for Microsoft-Decision-1 if you're already in the Microsoft stack — Foundry, Azure, Copilot — and want the decision model to live where your other infrastructure already is, backed by the largest vendor in the space.
• Reach for Jev when you want a hosted, RLCD-calibrated model with pinned versions and honest confidence, a drop-in API you can call today without Foundry onboarding, and a free browser playground to try one decision before you integrate.
• Because both speak the same shape — context plus a finite answer space in, a calibrated decision out — you can prototype on one and keep the other as a fallback or a provider you switch to later with little rework.

Make one real decision

The fastest way to judge any decision model is to run your own decision through it. Jev's playground does it in the browser with no signup, and a jv_live_ key points your code at the hosted endpoint — the same state-plus-typed-questions shape a decision-scoring model like Microsoft-Decision-1 expects:

// Hosted, calibrated decision — the same context + fixed-answer-space
// shape a decision model like Microsoft-Decision-1 expects.
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: { feedback: item.text },
    questions: {
      topic: { type: "choice", instructions: "Classify this feedback.",
               criteria: { bug: "something broken", request: "asking for a feature", praise: "positive note" } },
      ship_blocker: { type: "noul", instructions: "Is this a release blocker?" },
    },
  }),
});
const { answers } = await res.json();
bucket(answers.topic.choice, answers.ship_blocker.noul);

• Digital Today — Microsoft joins the decision-focused AI model race with Microsoft-Decision-1 — Task types, Qwen3.5-9B backbone, Foundry/OpenRouter availability, and the per-team speed/cost claims (vendor-reported).
• Crypto Briefing — Microsoft launches Decision-1, an AI model built to make fast calls — Nadella's announcement framing and positioning against other decision models.

FAQ

See also: What is a decision model · What is a decision API · Perplexity Decider vs Jev · OpenAI Decisions API vs Jev · Jev alternatives · Playground