Is Bev a newly trained model?

No. The README states the 7.21 GB GGUF is the original Prism file redistributed byte-for-byte, while Bev adds the serving integration, scoring extension, typed API, and evaluation tools.

Will it run on an 8 GB CPU-only machine?

Likely not. The digest says to start CPU-only use with 16 GB of RAM, calls 8 GB unverified and likely too tight, and notes RAM varies with context length and concurrent requests.

Do the 95% Persian scores mean Jev parity?

No. The comparison uses the benchmark author's published Jev reference rather than a new Jev run, on synthetic data; the report also lists weaker general diagnostics such as 7/12 MMLU and 2/10 SimpleBench.

Can I expose the API to the network?

You should not without a front end. It binds to loopback by default and has no built-in public authentication, and min_probability is only an abstention heuristic rather than a reliability guarantee.

Which install path should I pick?

The original Linux NVIDIA or CUDA installer runs scripts/install.sh and scripts/start-services.sh; macOS Apple Silicon and Linux or Windows CPU use the portable installer, python3 scripts/install-portable.py, then scripts/serve.py. Windows CPU is untested.

How does it answer without generating text?

It maps each allowed answer to a distinct single-token label and does one next-token scoring step per field, then normalizes those scores over the candidates; the typed JSON is assembled by the API.

Does it need a specific llama.cpp?

Yes. This release requires the pinned Prism fork of llama.cpp, and a generic GGUF loader is not sufficient; the original CUDA installer verifies the pinned model and runtime downloads before serving.

CommunityJev Demos & Experiments

Bev

Local HTTP API that scores 2-255 candidate options per field with a redistributed ternary GGUF, returning typed Choice, Noul, and Score answers as structured JSON.

See highlights

Open github.com

Stars: 12
Language: Python
License: MIT
Published by: Community
Author: Reza2kn
Last verified: Oct 2026

Reza2kn/Bev

User repository on GitHub

View on GitHub

Ternary decision scoring with Jevfire-style inference, typed APIs, and reproducible Persian evaluation.

Forks: 0
Last pushed: Updated Sep 2026

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

About this project

7.21 GB model file · 7.28 GiB measured CPU RAM · 8.18 GiB observed Mac Metal process RSS · 8.30 GiB observed NVIDIA VRAM. Bev runs Jevfire-style decision scoring on Prism ML's Ternary-Bonsai-2-27B. Give it context and a finite set of choices; it scores every candidate, selects an answer, and returns structured JSON through a local HTTP API.

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

Tagscommunity jevbest catalog-import

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Python

Highlights

One next-token scoring step per field over allowed labels, with candidate probabilities and token IDs returned.
Serves on loopback 127.0.0.1:18781 with interactive docs; data dir defaults under home and BEV_ROOT overrides it.
Requires the pinned Prism llama.cpp fork and the 7.21 GB model file; CPU-only hosts should start with 16 GB RAM.
Fields evaluate independently; Noul returns P(true) and Score returns a probability-weighted rubric position.

Watch out

Persian benchmark used the Linux CUDA path on synthetic data and does not establish general reliability; scores are relative preferences, not calibrated confidence. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

open-jev-typed-decision-engine

Open 150M encoder that reproduces the typed noul, choice, and score interface in one forward pass, trainable on a free GPU and runnable locally for nothing.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

intikhab49/open-jev-typed-decision-enginePython

TypeLLM

941

Library that adds schema-guaranteed typed outputs to existing autoregressive LLMs without changing weights, served through SGLang or a hosted API.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

TypeLLM/TypeLLMPython

solar-mini4-jev

Drop-in service that exposes Upstage Solar models through the System One shape, accepting state plus noul, choice, or score questions.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

hunkim/solar-mini4-jevPython

open-medical-jev

Research recipe that fuses two frozen open readers into Jev-style yes/no judgments with a fit-free router, confidence gate, and candidate sets.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

FeiLiuEM/open-medical-jevPython

jev-robot-control

Simulation study that drives one xArm7 apple-to-plate task with Jev, a large reasoning model, and a small model through identical incremental controls.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

openroboto-ai/jev-robot-controlPython

OpenJev-Vision

Open research toolkit that encodes an image once and answers multiple typed choice, noul, and score questions from a shared distribution, plus a local 0.6B text-decision model.

Jev Demos & Experiments#community#jevbest#catalog-import

Community

IamBusy/OpenJev-VisionPython

From the Jev guides

Original write-ups on what this entry is for.

Per-Tick Decisions in Games

Sep 28, 2026

A game tick is a state plus a legal action list. How Clash Royale, Minecraft, and the earlier Mario, Pokémon, StarCraft, and drone loops ask Jev one step at a time.

Jev Demos & Experiments#gaming#agents#control

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai