Jev directory
CommunityJev Models & Reimplementations
N-Q-26M-JEV-1027
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.
- Reading time
- 2 min
- Published by
- Community
- Added
- Oct 2026
Highlights
- 46,453,084 parameters total: 19 I64 and 46,453,065 F32 tensors, in a 185,827,092-byte file.
- Loads via Transformers AutoModel with the custom class modeling_nq4.NQ4ForDecision.
- Repo tags include decision-model, nli, calibrated, distillation, gated-deltanet, and tiny-model.
- Training data lists facebook/xnli, stanfordnlp/sst2, fancyzhx/ag_news, and IlyaGusev/saiga_preferences.
- Chat template uses <bos>, <SYSTEM>, <USER>, <MODEL>, <EOT>, and an optional <THINK> marker.
Questions about this entry
Short answers drawn from the linked project or page.
The model card shows a text-classification pipeline pointing at q-project/N-Q-26M-JEV-1027 and an AutoModel entry with the custom class modeling_nq4.NQ4ForDecision. Because the repo carries the custom_code tag, the runtime has to load that custom modeling file rather than a stock architecture.
The safetensors metadata reports 46,453,084 parameters in total, split into 19 I64 tensors and 46,453,065 F32 tensors. The single unsharded file weighs 185,827,092 bytes, and the model is not marked as quantized.
The model card lists two languages, Russian (ru) and English (en), and also carries a russian tag. The sample widget text shown on the page is Russian, and the training mix includes both Russian and English datasets.
The repository is tagged license:apache-2.0 and is not gated, so the weights can be downloaded without an access request. The model card includes a LICENSE file path, and the repo is public under the q-project organization.
Tags name AlexWortega/openjev-data, IlyaGusev/saiga_scored, IlyaGusev/saiga_preferences, HuggingFaceH4/ultrafeedback_binarized, facebook/xnli, MonoHime/ru_sentiment_dataset, stanfordnlp/sst2, s-nlp/ru_paradetox_toxicity, and fancyzhx/ag_news, spanning NLI, preferences, sentiment, news, and toxicity.
The page does not state any link to TypeSafe AI. It is published under the q-project organization on Hugging Face and is described as a finetune of the base model q-project/N-Q-26M-JEV, created and last modified on 2026-10-27.
It is a tiny 46M-parameter text-classification model tagged decision-model and tiny-model, not a general chat or generation system, and no inference providers are listed for it. Treat it as a small calibrated classifier for Russian and English labels rather than a broad language model.
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
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 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
MLX builds of Jev-Style-2B-Decision-v4 for Apple silicon: bf16 (3.76 GB) and 8-bit (2.00 GB) folders for text and JSON states, run through one shared jev_style_decision_mlx.py runtime under Apache-2.0.
Jev Models & Reimplementations#open-models#calibration#local
Community
A 26M-class decision model on Hugging Face, fine-tuned from q-project/N-Q-26M-Chat. It ships a custom NQ4ForDecision class for text classification and feature extraction, is tagged calibrated and decision-model, and publishes Apache-2.0 weights with Russian and English support.
Jev Models & Reimplementations#open-models#calibration#classification
Community
Encoder-based typed decision engine from ScaiResearch Labs: a frozen ModernBERT-base backbone with rank-16 LoRA adapters that returns one answer per typed question as a calibrated distribution over the supplied options. Ships a training checkpoint and Kieve inference config.
Jev Models & Reimplementations#open-models#calibration#classification
Community
An EmbeddingGemma 2 fine-tune that answers choice, score, and noul questions from a state and a plain-data schema, with a 0.5M-parameter schema-conditioned scorer over 768-d embeddings.
Jev Models & Reimplementations#community#open-models#local
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.