User repository on GitHub
非自回归System 1决策引擎，专为结构化类型决策场景设计 DeepOpen Multilingual, non-autoregressive System 1 decision engine.
- Forks
- 220
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.Homepage
CommunityJev Models & Reimplementations
Deepopen
DeepOpen is a non-autoregressive decision engine built on Laya that classifies into developer-defined types in one forward pass, with a language router picking between English, multilingual, and typed-decision checkpoints.
- Stars
- 2k
- Language
- Python
- License
- APACHE-2.0
- Published by
- Community
- Author
- deepopen-com
- Last verified
- Oct 2026
About this project
Open-Source Multilingual System 1 Decision Engine Technical Whitepaper.
Quoted from the project's README. Read the full README.
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- Python
Highlights
- Questions are typed as choice, score, or noul, so output is a predefined label with a probability instead of generated text.
- A pure-Python script and language check runs in under 0.5ms to pick a checkpoint before inference.
- Banking77 and CLINC150 reproduction scripts, dataset configurations, and pretrained checkpoints are in the repository.
- Apache 2.0 with all weights open; pip install deepopen and Router(preload=True, device="cuda") is the entry point.
Watch out
The comparison table against a hosted decision API is published by this project rather than a third party, and the README's own latency figures conflict: the English section says 7.2ms for batch processing while the Chinese section reports 72.3ms for ten questions. Hardware results are stated for an NVIDIA T4. 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.
The README says a pure-Python detection step runs in under 0.5ms before inference to pick a checkpoint, because the English-only checkpoint scored 0 accuracy on a Khmer task while still reporting 95.2% confidence. A confidence threshold cannot catch that, so routing does.
Not without changing a parameter. The README concedes the API it benchmarks against performs better in high-cardinality cases with 50+ options, and that DeepOpen needs head_max_len raised to 512 to push its option prompt budget to support 70+ choices.
No. The comparison table is published by this project. What is independently reproducible is the intent-classification side: full reproduction scripts, dataset configurations, and checkpoints for Banking77 and CLINC150 are included in the repository.
Router(preload=True, device="cuda") loads everything. You can pass a list to router.preload(["english", "multilingual"]) to load only the checkpoints your business needs, cap resident models with Router(max_loaded=2), and free them with router.unload().
An English checkpoint (ModernBERT-large, 421M parameters, 512-token window), a multilingual one (mmBERT-base, 322M parameters, 1024 tokens, 100+ languages), and a typed-decisions checkpoint (ModernBERT-large, 1024 tokens) tuned for structured decision scenes.
0.766 on a 2,000-sample structured-decision benchmark, above TypeSafe Jev 1.13.0's 0.727, with a Brier score of 0.062 and ECE of 0.081 after domain temperature calibration.
Yes. The README links a Kaggle notebook that runs the full fine-tune on two free T4 GPUs in 4 to 5 hours over 30,000 questions, and says RLCD raised base accuracy from 0.36 to 0.766.
More Jev resources like this
1.7kGitHub stars
Open-weight 4B and 9B models from TokenRhythm: NeoHorse-1 for tool use, coding, and instruction following, plus NeoHorse-Jev-4B for prefill-only Choice, Noul, and Score decisions.
Jev Models & Reimplementations#community#jevbest
Community
TokenRhythm/NeoHorsePython
TokenRhythm/NeoHorsePython1.5kGitHub stars
A 0.8B decision model that answers first and escalates only when unsure, plus nine LoRA adapters of roughly 41 MB each for guard, triage, tool choice, spam, and similar classification work.
Jev Models & Reimplementations#community#jevbest
Community
firelex/jeffPython
firelex/jeffPython715GitHub stars
Decision models that answer several choice, score, and noul questions over a state with probabilities in one forward pass, with training, two inference backends, calibration tooling, and the benchmark suite in the repo.
Jev Models & Reimplementations#community#jevbest
Community
InternLM/Intern-DecisionPython
InternLM/Intern-DecisionPython402GitHub stars
Decision models that return typed probabilities for candidate labels from a state, instructions, and options, shipped as LoRA adapters plus a trained decision head over pinned Qwen base weights.
Jev Models & Reimplementations#community#jevbest
Community
Zefan-Cai/Open-JevPython
Zefan-Cai/Open-JevPython363GitHub stars
Open-weight 2B, 4B and 9B decision models that extend Jev's typed question format to photos, so an image plus a record can be checked against options you supply, with a probability on each and an explicit can't-tell. Runs on your own hardware.
Jev Models & Reimplementations#community#jevbest
Community
mohit67890/imajevPython
mohit67890/imajevPython253GitHub stars
A clone of browser-use's jev-ultrafast that swaps the hosted Jev decision call for Laya running locally through MLX, keeping the browser agent, action space and executor but making no paid decision API calls. Apple Silicon only.
Jev Models & Reimplementations#community#jevbest
Community
ipenywis/laya-ultrafastPython
ipenywis/laya-ultrafastPythonFrom the Jev guides
Original write-ups on what this entry is for.
Jev takes text only; Clef-omni adds image, audio, and video input and Liquid d1 adds images, each keeping the same typed-answer shape. Where that matters, where it does not, and the costs the vendors underplay.
Jev Models & Reimplementations#multimodal#open-models#decision-model
Read article
One model in September, a dozen by October. Clef, Liquid d1, Perplexity, and OpenAI all ship the same typed-answer shape now. How to read the rival benchmarks and pick on your own task.
Jev Models & Reimplementations#decision-model#benchmarks#pricing
Read article
Hundreds of repositories carry the system-one tag, but the ecosystem is a dozen ideas in different sizes. A tour of the families, the licenses, and what a download does not include.
Jev Models & Reimplementations#open-models#hugging-face#local
Read article
From the Jev community
Posts from builders shipping with Jev right now.