Perplexity Decider vs Jev

Perplexity's Decider is pplx-decider-v1-27b — an open-weight decision model behind a new Decisions API that Perplexity launched on 1 October 2026. Like Jev, it takes application state plus typed questions and returns a calibrated probability for each option instead of free-form text. It arrived the same week as Cloudflare's Clef, making it the second serious open-weight challenger in the category Jev defined. Here's the honest comparison — specs, Perplexity's own benchmark claims, and when to pick which.

If you're weighing Perplexity's Decider against Jev, both answer the same question in the same shape: a class, a yes/no, or a rating with a probability attached, in one pass. The choice is about open-weight versus hosted-and-calibrated — not about whether the model can decide.

What pplx-decider is

Perplexity's Decisions API runs on pplx-decider-v1-27b, an open-weight model fine-tuned from Qwen3.8-27B (about 26B parameters) and released on Hugging Face under Apache-2.0. Instead of writing text, it picks one option from a labelled set and returns calibrated probabilities, or a yes/no probability for a question; it's multimodal, taking images alongside text. Perplexity frames the typical use as support triage — the Decisions API makes the first call on each ticket, and only escalations go to its heavier Agent API. A newer checkpoint, pplx-decider-v1.1-27b, followed around 7 October.

Specs, side by side

The pplx-decider column uses Perplexity's published figures; the Jev column uses TypeSafe's. Every number here is vendor-reported.

Perplexity's benchmark claim — read it honestly

Perplexity reports a composite accuracy of 85.71% for pplx-decider across an 11-task panel, ahead of Jev's 84.51% and the Qwen3.8-27B base's 74.76%. The headline is a narrow edge — but the same breakdown shows Jev scoring higher on 6 of the 11 individual tasks, so the composite win rides on a few tasks, not a uniform lead. These are Perplexity's internal numbers, not an independent third-party evaluation.

That makes the honest read 'competitive, roughly a wash on accuracy by the vendor's own panel' rather than 'beats Jev'. As always, the only benchmark that settles it is your own task on your own data — a composite from either vendor won't tell you which wins the specific decision your app makes.

When to pick which

• Reach for pplx-decider when you want open weights and the lowest sticker price per input token — to self-host, to keep decisions inside an existing Perplexity stack, or to feed very large context (its advertised window is far bigger than Jev's).
• Reach for Jev when you want a hosted, managed model with RLCD-calibrated confidence and pinned versions, no inference ops to run, and a drop-in API plus a free browser playground to try before you integrate.
• Because both speak the same decide shape — state plus typed questions, probabilities out — you can prototype on one and keep the other as a fallback 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:

// Hosted, calibrated decision — the same state+questions shape an
// open-weight decider like pplx-decider 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: { ticket: ticket.body },
    questions: {
      triage: { type: "choice", instructions: "First-pass triage.",
                criteria: { resolve: "answerable now", escalate: "needs a human", spam: "not a real request" } },
      urgent: { type: "noul", instructions: "Is it urgent?" },
    },
  }),
});
const { answers } = await res.json();
if (answers.triage.choice === "escalate") handoff(ticket, answers.urgent.noul);

• Hugging Face: pplx-decider-v1-27b — Perplexity's open-weight model card — Qwen3.8-27B fine-tune, Apache-2.0, 11-benchmark table (vendor-reported)

FAQ

See also: What is a decision model · What is a decision API · Cloudflare Clef vs Jev · Jev alternatives · Playground