Jev directory
CommunityJev Models & Reimplementations
JEV-27B for Core AI
A Hugging Face bundle that ports AutoTrust AI's JEV-27B to Apple's Core AI framework, with a Swift command-line runner for typed yes/no, multiple-choice, and rating decisions on Apple silicon Macs.
- Reading time
- 2 min
- Published by
- Community
- Added
- Nov 2026
Highlights
- Runs JEV-27B locally on Apple silicon Macs through Apple's Core AI framework and a Swift CLI runner.
- Takes a situation plus a yes/no, multiple-choice, or rating question and returns a probability for each allowed answer.
- Model bundle is 13 files, 16,064,821,387 bytes (14.96 GiB), quantized to uint4 with the qwen3_5_cfg recipe.
- Measured on M4 Max 128 GiB: 0.615 s for a 100-token decision, 8.978 s for 2,000 tokens, 37.139 s for 8,192 tokens.
- Requires Apple silicon, macOS 27, Xcode 27 with Metal Toolchain, Rust/Cargo, and the Hugging Face CLI.
Questions about this entry
Short answers drawn from the linked project or page.
Download the bundle with the Hugging Face CLI, clone the oka-yuji/coreai-samples repository, run Tools/Irodori/build-inputs.sh, then build the coreai-verify product with swift build -c release and call coreai-verify typed-decide with the bundle path, a request.json input, and an output path.
Apple silicon, macOS 27, Xcode 27 with the Metal Toolchain, Rust/Cargo, and the Hugging Face CLI. Execution is GPU-based and was measured on an M4 Max with 128 GiB RAM. Minimum memory and sustained-use limits are not established.
You give it a situation plus a yes/no, multiple-choice, or rating question, and it returns a probability for each allowed answer. It is not built for free-form chat, iOS, or high-risk decisions without independent validation.
On M4 Max 128 GiB with batch size and concurrency 1, medians of three warmed runs were 0.615 s at 100 tokens, 2.208 s at 500 tokens, 8.978 s at 2,000 tokens, and 37.139 s at 8,192 tokens. Decision time excludes loading and JSON file I/O.
The model card lists apache-2.0. The bundle ships LICENSE and NOTICE.md files covering license, attribution, and file-level modification, and bundle/SHA256SUMS holds checksums for every other file in the bundle.
Minimum memory and sustained-use limits are not established, and footprint kept increasing during repeated requests of alternating lengths. The recorded example is a usage check, not a latency benchmark, and no controlled speed comparison with the original model or Core ML is claimed.
No. The instructions say to run one model job at a time, and the companion repository documents its shared model-job lock in AGENTS.md. The output file must not already exist when you invoke the runner.
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 NewsHacker 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 NewsMore Jev resources like this
LoRA adapter for Qwen3.5-4B that scores medical multiple-choice options in a single forward pass and returns a short JSON object with the chosen answer key and per-option probabilities, without explanations or chain-of-thought.
Jev Models & Reimplementations#open-models#python#classification
Community
Satirical Python script that answers a query with DO or DO_NOT plus a random confidence percentage, presented as a natural alternative to Jev. MIT-licensed, Python 3.8+, one dependency, no API keys or model weights.
Jev Models & Reimplementations#github#python#open-models
CommunityAhmad170412/Koin
Russian-first open typed-decision model on Hugging Face. Given a context and a set of candidate options it returns a calibrated probability and picks one, covering yes/no, choice, and ordinal scale decisions. Built on mmBERT-base with 307M parameters under an MIT license.
Jev Models & Reimplementations#open-models#calibration#classification
Community
A 561M-parameter XLM-RoBERTa-large decision model given one extra training epoch on Jev's answers, with about 23% of examples black-box distilled from Jev's public API. Supports noul, choice, and score questions in 18 languages and loads through the bundled naluri.py.
Jev Models & Reimplementations#open-models#calibration#benchmark
Community
A trained example bank for @ruvector/typesafe on Hugging Face, covering 150 CLINC150 intents across 10 domains plus 1,000 out-of-scope utterances. It stores labelled examples and frozen splits, not weights; heads and temperature calibration are refit on load.
Jev Models & Reimplementations#open-models#classification#local
Community
A 46M-parameter Apache-2.0 decision model on Hugging Face, fine-tuned from q-project/N-Q-26M-JEV for typed classification. It exposes the custom NQ4ForDecision class to Transformers and lists Russian and English, plus NLI, sentiment, news, and toxicity datasets.
Jev Models & Reimplementations#open-models#classification#calibration
Community
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.