Jev directory
CommunityJev Models & Reimplementations
N-Q-26M-JEV
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.
- Reading time
- 2 min
- Published by
- Community
- Added
- Oct 2026
Highlights
- Safetensors checkpoint totals 46,453,084 parameters at roughly 186 MB, unsharded and not quantized
- Custom architecture: the card lists custom class modeling_nq4.NQ4ForDecision under AutoModel
- Tagged decision-model, calibrated, gated-deltanet, tiny-model, nq4, and custom_code
- Model card lists eight datasets, including stanfordnlp/sst2, facebook/xnli, fancyzhx/ag_news, and IlyaGusev/saiga_scored
- Licensed Apache-2.0; card lists Russian and English, with 3 likes and 0 downloads at listing time
Questions about this entry
Short answers drawn from the linked project or page.
The model card shows two paths: a Transformers pipeline call with task text-classification and model q-project/N-Q-26M-JEV, or loading directly with AutoModel.from_pretrained using device_map="auto". Both snippets pass trust_remote_code=True, and the card notes installing or upgrading transformers and accelerate.
Yes. The repository carries the custom_code tag and its transformersInfo entry names the custom class modeling_nq4.NQ4ForDecision, so the documented loading snippets set trust_remote_code=True. Expect to review the bundled modeling code before running it in a restricted environment.
The model card lists license: apache-2.0 and the repository is not gated, so the weights can be downloaded without an access request. The card also links a LICENSE file in the repository for the full terms.
The card lists Russian (ru) and English (en) as languages, and carries a russian tag alongside the nq4 and tiny-model tags. The sample widget text shown on the page is Russian, and several of the listed training datasets are Russian-language sets.
The card records base_model:q-project/N-Q-26M-Chat with the finetune relation, so this checkpoint is presented as a fine-tune of the Q Project N-Q-26M-Chat model rather than a from-scratch training run.
The safetensors metadata reports 46,453,084 parameters (46,453,065 in F32 and 19 in I64) with a total file size of about 185.8 MB. The checkpoint is not sharded and the repository is marked as not quantized.
Eight datasets are tagged: 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.
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
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
Hugging Face repository hosting GGUF weights for Jev-Style-2B-Decision-v4, a 2B Jev-style decision model published under the chaoliangUNSW account. The listing includes a chat template that renders messages, reasoning content, tool calls, and vision inputs.
Jev Models & Reimplementations#open-models#local#calibration
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
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
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.