Is it really wire-compatible, or will my Jev SDK break?

The server speaks Jev's wire format and the SDK works unchanged for yes/no questions. One wrinkle: on a noul, Jev's optional criteria field is accepted but not read, so policy text has to go into the instructions instead.

How much accuracy do I give up versus hosted Jev?

On 2,000 yes/no items from policies it never saw, the project reports 0.810 against Jev's 0.927 and Laya's 0.489. On five public tasks it is behind Jev across the board, for example 0.75 versus 0.90 on medical Q&A. These are the project's own comparisons.

Can it do choice and score, or is it yes/no only?

The sources disagree. The README documents choice and score in Jev's shapes, with scores marked early; the listing records both as refused with HTTP 422 until they ship. Under the hood a choice becomes one yes/no question per option, so check the release notes for the build you install.

What does it cost to run, and how fast on a laptop?

One binary, 8-bit weights through OpenVINO, CPU only, no Python and no GPU; GGUF weights also ship for llama.cpp. The current README medians are 26 ms for a short request and 112 ms for a long one on an Intel Core Ultra 7 with 16 threads and the state cache off; a cached state brings repeats to 22 ms.

How do I start the jevos server locally?

Download the archive for your system plus the openvino-int8 model zip from the release, then run ./jev serve (jev.exe on Windows); it listens on 127.0.0.1:8017 and speaks POST /v1/systemone.

Do I have to run a server to use jevos?

No. ./jev decide request.json answers a request file without starting the server, and the model also ships as GGUF for llama.cpp, so the same weights can run through another runtime.

How do I protect a jevos server?

With JEV_API_KEY set, every call except /health needs an Authorization: Bearer header; it uses about 1 GB of memory, or 1.4 GB with the text cache full, and --threads controls CPU use.

Jev directory

CommunityJev Models & Reimplementations

Jevos

A local yes/no decision model that speaks Jev's wire format on a laptop CPU, refusing Choice and Score until those question types ship.

See highlights Open github.com

Stars: 1.4k
Language: C++
License: MIT
Reading time: 3 min
Published by: Community
Author: feder-cr
Last verified: Sep 2026

feder-cr/jev

User repository on GitHub

View on GitHub

jevos: an open-source alternative to Jev and Microsoft-Decision-1. Yes/no decisions, choices and scores with probabilities, on your laptop's CPU.

Forks: 167
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

Decisions on a laptop CPU in 28–130 ms. Send a text and a question (yes/no, multiple choice or a score) and get back the probability of each answer, in TypeSafe Jev's API, on your own machine. From the release, download the archive for your system and jevos-v4-openvino-int8.zip, then: The model is also on Hugging Face, feder-cr/jev: hf download feder-cr/jev --include "model/" --local-dir jev puts it in jev/model/.

Quoted from the project's README. Read the full README.

Tagscommunity python open-models local noul wire-compatible decision-server

Works forJev for Classification Jev for Decision gating

Used inJev for Software & DevTools

Languages: Python
Platforms: API CLI
Capabilities: Open sourceSelf-hostedRuns locally

Highlights

Yes/no only: choice and score questions are refused with HTTP 422.
On 2,000 unseen yes/no items the README reports accuracy 0.815, against 0.927 for Jev and 0.489 for Laya.
Short requests are listed at 54 ms on CPU; three questions sharing one state take about 165 ms.
Context is 8,192 tokens. Optional noul criteria are accepted and not read.
A bundled /dino page plays a Chrome Dino-style game from two yes/no questions per step.

Quickstart

bash

uv run jev serve --gguf jevos-q4_k_m.gguf --device cpu --threads 16
curl http://127.0.0.1:8017/v1/systemone -H 'Content-Type: application/json' -d '{
  "model": "jev-latest",
  "state": "I was charged twice for the same order.",
  "questions": {"billing": {"type": "noul", "instructions": "Is this a billing problem?"}}
}'

Watch out

MIT-licensed. Independent of TypeSafe. Multiple choice and scores are marked soon, and the published accuracy is the project's own comparison.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

X: Standard One SH (StandardThinking)

LinkedIn: What multimodal decision models actually change, and what they do not.

LinkedInJevDirectory.org

Hacker News: Decision Index 0.2.1 scores open reproductions of Jev on a frozen suite.

jackbravo4 pointsDecision Index 0.2.1 – 70 open reproductions of TypeSafe Jev's Decision ModelRead the thread on Hacker News

Hacker News: Pointer to a DiffusionGemma plus vLLM tutorial for a Jev-shaped classifier.

1 pointHow to build a Jev-style classifier with DiffusionGemma and vLLMRead the thread on Hacker News

From the Jev guides

Original write-ups on what this entry is for.

Serving System One Locally

Sep 28, 2026

A base URL is the whole client change. What still works when Ollaya, Lichen, verdict, Rizzo Flow, or jevos answers POST /v1/systemone on your machine.

Jev Models & Reimplementations#open-models#local#api

Read article

The Open Decision-Model Roundup

Oct 14, 2026

Training recipes, looped architectures, and language ports: what the open side of the typed-decision class shipped this month, and how to pick by runtime and calibration instead of the leaderboard.

Jev Models & Reimplementations#open-models#training#local

Read article

Exporting Decision Models: GGUF, ONNX, and Device Runtimes

Oct 10, 2026

Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.

Jev Tools & CLIs#open-models#local#onnx

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai

Jevos

About this project

Highlights

Quickstart

Questions about this entry

Is it really wire-compatible, or will my Jev SDK break?

How much accuracy do I give up versus hosted Jev?

Can it do choice and score, or is it yes/no only?

What does it cost to run, and how fast on a laptop?

How do I start the jevos server locally?

Do I have to run a server to use jevos?

How do I protect a jevos server?

Reactions & coverage

Related terms

More Jev resources like this

From the Jev guides

From the Jev AI model community