Jev directory
CommunityJev Models & Reimplementations
N-Q-26M-JEV-1110
A 46M-parameter Russian and English decision model published on Hugging Face, tagged for classification, NLI, toxicity, moderation, and calibrated distillation. It ships a custom NQ4ForDecision class and a chat template with an optional thinking block.
- Reading time
- 2 min
- Published by
- Community
- Added
- Nov 2026
Highlights
- 46,453,084 parameters stored as F32 safetensors, roughly 186 MB total file size.
- Custom architecture loaded through AutoModel with custom class modeling_nq4.NQ4ForDecision.
- Tagged decision-model, nli, toxicity, moderation, profanity, and calibrated distillation.
- Fine-tuned from q-project/N-Q-26M-JEV-1027 and listed for Russian and English.
- Chat template uses <SYSTEM>, <USER>, <MODEL>, <EOT>, and an optional <THINK> block.
Questions about this entry
Short answers drawn from the linked project or page.
The repository uses custom code, so loading requires the transformers library with remote code enabled and the custom class modeling_nq4.NQ4ForDecision, exposed through AutoModel. Weights are stored in safetensors format and the pipeline tag is text-classification, with feature-extraction listed as the transformers auto-model task.
The model card lists two languages: Russian (ru) and English (en). Its tags include a russian label, and the sample widget text shown on the page is Russian, so Russian-language classification and moderation inputs are the primary documented surface.
The safetensors index reports 46,453,084 total parameters, split into 46,453,065 F32 values and 19 I64 values, with a total file size of 185,827,092 bytes. That is small enough for local inference, and the repository is not marked as quantized.
The repository carries the license tag "other" and includes a LICENSE file at the root of the model repo. Because the license is not a standard identifier, you should read that file directly before redistributing weights or using the model commercially.
Tags include decision-model, classification, nli, toxicity, moderation, profanity, calibrated, distillation, custom-architecture, gated-deltanet, and tiny-model. The Hugging Face pipeline tag is text-classification, and the transformers info lists feature-extraction as the auto-model task.
The card references AlexWortega/openjev-data, AlexSham/Toxic_Russian_Comments, textdetox/multilingual_toxicity_dataset, google/civil_comments, 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.
It is a tiny distilled model with 2 all-time downloads and 1 like at the time of listing, and the page publishes no model-index or benchmark results. Treat its calibrated probabilities as unverified for your domain and measure them on your own labeled data before relying on its outputs.
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
v-jev 1.4 is a Jev-style decision model on Hugging Face in 0.8B and 4B sizes, built on Qwen3.5 backbones. It returns probabilities for typed questions in one forward pass and adds value extraction and image decisions.
Jev Models & Reimplementations#open-models#classification#calibration
Community
A 7.7M-parameter decision model on Hugging Face that reads text or speech and returns one probability per option for choice, score, and noul questions. Speech runs through femto-asr, giving a 10.1M-parameter spoken pipeline.
Jev Models & Reimplementations#open-models#classification#local
Community
Fine-tuned UltraBERT-380M decision model on Hugging Face that scores a state, question, and option list in one forward pass. Uses sigmoid of yes minus no normalized over options; 72.8% dev and 71.9% test accuracy across 13 datasets, with an evidence head documented on the card.
Jev Models & Reimplementations#open-models#arxiv#calibration
Community
A Hugging Face checkpoint that fine-tunes UltraBERT-380M into a Jev-style decision model: it reads a state, a question and a list of options and picks one in a single forward pass. Reports 72.6% mean accuracy over 13 datasets on the test split.
Jev Models & Reimplementations#open-models#classification#local
Community
Fine-tuned UltraBERT-380M decision model on Hugging Face that reads a state, question, and option list and picks one option in a single forward pass. Scores options with a square-root softplus of yes minus no, normalized over options; 73.2% dev and 71.9% test accuracy over 13 datasets.
Jev Models & Reimplementations#open-models#arxiv#calibration
Community
A trained example bank for @ruvector/typesafe holding 9,993 labelled Banking77 utterances across 77 banking intents with frozen split assignments. Heads and temperature calibration are refit from the bank on load; no weights are stored.
Jev Models & Reimplementations#open-models#classification#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.