Jev on Ollama

If you're looking to 'ollama pull jev', the honest answer is that hosted Jev is a closed, calibrated model with its own endpoint — it isn't an Ollama model. But the thing you actually want — a typed decision model running locally — does exist: Ollaya is the 'Ollama for decision models', and it serves open Jev-like models behind a TypeSafe-compatible API.

Jev (TypeSafe) is hosted and closed-weight, so there's no GGUF to pull into Ollama. What the community built instead is Ollaya — a local runtime that pulls and serves open decision models (Laya, winnow, NLI, GLiClass and more) by name, the way Ollama serves LLMs. It speaks TypeSafe's /v1/systemone wire format, so code written for Jev points at it by changing one base-URL env var. These are community, open-weight models — not Jev, and not calibrated the way hosted Jev is — so validate any of them on your own data.

Run one locally

# Install Ollaya (the 'Ollama for decision models')
curl -fsSL https://ollaya.dev/install.sh | sh

# Pull + serve an open Jev-like model (Laya = ModernBERT-large, 421M, CPU-friendly)
ollaya run laya --preset triage
# GGUF models (winnow, jevk5) run on llama.cpp (CPU/CUDA/Metal); on Mac, laya runs on the Apple GPU via MLX

Ollaya exposes a TypeSafe-compatible endpoint locally, so the same decide call you'd send Jev works against it — you just point the base URL at the local daemon:

// Same request shape as hosted Jev — just a local base URL, no key
const res = await fetch("http://localhost:11435/v1/systemone", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({
    state: { ticket: text },
    questions: {
      queue: { type: "choice", instructions: "Which queue should this go to?",
               criteria: { billing: "payments", technical: "bugs", abuse: "spam/fraud", sales: "pricing" } },
    },
  }),
});
const { answers } = await res.json();
route(answers.queue.choice); // one of your criteria keys — same shape Jev returns

What you give up running locally

• Calibration — open models aren't RLCD-calibrated, so a 0.8 doesn't reliably mean 80%; you own the validation
• Accuracy — the small open models trade points of accuracy for size and speed; check them on your labelled data
• Pinning & ops — you run the daemon, pick models, and keep them updated; nothing is versioned for you
• No confidence you can trust a production threshold against out of the box

When to run local vs call hosted Jev

The clean split: run Ollaya locally for offline, private, or high-volume bulk decisions where you own the box, and call hosted Jev when you want a calibrated, pinned decision you can trust a threshold against with no ops. Both answer the same typed shape — choice, score, noul — so the interface your code depends on stays identical. See the local-model guide for the full self-host path.

FAQ

See also: Run a decision model locally · Laya vs Jev · Jev alternatives · Pricing