CommunityJev Models & Reimplementations
Julia 1
Supersonic Labs' 144.3M-parameter decision model on mmBERT-small: choice, score, and noul answers from supplied options in one pass, 73.15% on 2,000 typed decisions, Apache-2.0.
- Published by
- Community
- Author
- Supersonic Labs
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- 144.3M parameters with a 550.5 MiB fp32 checkpoint, trained from JHU CLSP's mmBERT-small encoder.
- One pass scores all named questions; each choice accepts 2 to 20 options and score answers are zero-based rubric indexes.
- Its 73.15% on typed decisions edges the supplied Jev reference (72.70%) by 0.45 points, per the card's own protocol.
- The Banking77 pilot trails at 64/100 against an 87/100 reference — the card's own example of where a small model misses.
Quickstart
python
from julia import load_model
engine = load_model("Julia-1", device="cpu", strict_encoding=True, max_length=8192)
result = engine.predict(
state="I was charged twice for the same order.",
questions={
"team": {
"type": "choice",
"instructions": "Which team should handle this request?",
"criteria": {"billing": "Billing disputes", "shipping": "Delivery issues"},
},
},
)
print(result["answers"]["team"]["choice"])Watch out
Native calls accept 2–20 options. Grouped routing can lose the correct answer while narrowing, and its final probabilities cover only the finalists, so a grouped result is not a global distribution. The 73.15% comes from the checkpoint's own benchmark protocol, not a third-party rerun.
Questions about this entry
Short answers drawn from the linked project or page.
No. It scores supplied answers and returns the selected option with full softmax probabilities; it does not generate text. The card says it cannot be counted on to supply missing facts, solve equations, or carry long chains of reasoning just because the right answer is among the options.
The native call accepts 2–20 options. A hierarchical Router narrows groups and reranks survivors for larger lists, but the card warns that narrowing can drop the correct answer and that a grouped result is not a global probability distribution.
Load it from the julia package: load_model('Julia-1', device='cpu', strict_encoding=True, max_length=8192), then call predict with a state and a dict of typed questions. The card pins those runtime settings as the exact call.
On its typed-decision suite Choice reaches 71.33% (428 of 600), Noul 80.67% (484 of 600) and Score 68.88% (551 of 800), with abstentions counted as incorrect.
On MASSIVE scenario classification it reaches 71.50% macro accuracy across all 52 locales, 86.25% on Portuguese pt-PT and 86.75% on English en-US, built on a multilingual ModernBERT encoder. That suite does not measure intent classification or slot filling.
The runtime supports up to 8,192 combined tokens, though the historical benchmarks used 1,024. The upstream mmBERT-small architecture supports 8,192, and the released checkpoint keeps that context window.
No. They are supplied comparison references rather than a new Jev run. The Banking77 pilot uses a 72-label ranking with a top-16 shortlist rather than a native 72-option call, and sample sizes and prompts differ between the two.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
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 NewsHacker News: A one-token logprob wrapper that also classifies webcam frames.
allanrbo151 pointsA single function Jev-like wrapper for LLMs, including vision modelsRead the thread on Hacker NewsMore Jev resources like this
The Hugging Face hub for Laya's open System 1 checkpoints: the 421M English model at the repo root, a 322M multilingual build in a subfolder, and the typed-decisions fine-tune, all Apache-2.0.
Jev Models & Reimplementations#community#open-models#local
Community
Liquid AI's 3.12B decision model built on LFM2.5-VL: state plus typed questions in one forward pass with zero output tokens, 48.57 on Decision Index 0.2.1 and 8 ms per decision on an RTX 4090.
Jev Models & Reimplementations#community#open-models#local
Community
An untrained readout: stock Gemma-4-31B weights plus a decider config that reads option letters at the answer slot, scored #2 of 70 on Decision Index v0.2.1 at 57.33 against Jev 1.13.0's 57.91.
Jev Models & Reimplementations#community#open-models#local
Community
Nace.AI's 8B decision model on NVIDIA's Efficient-DLM diffusion backbone: a 256-dim pointer head scores every option in one pass, behind the same POST /v1/systemone API as its hosted service.
Jev Models & Reimplementations#community#open-models#local
Community
The vLLM Semantic Router project's 29.37B decision model on Qwen3.8-27B: a probability per supplied option with zero generated tokens, 74.0 on JevArena and 56.5 on Jev Decision Index 0.2.1.
Jev Models & Reimplementations#community#open-models#local
Community
A 4B scorer rebuilt from Qwen3-4B into a branching cross-encoder: one branch per option, no position bias, up to 256 options, and calibrated probabilities with no text to parse.
Jev Models & Reimplementations#community#open-models#local
Community
From the Jev guides
Original write-ups on what this entry is for.
Interface, size, license, claims — in that order. A selection guide across the system-one catalog, from a 706K form scorer to 27B Jev students.
Jev Practices & Patterns#open-models#evaluation#benchmarks
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
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
From the Jev community
Posts from builders shipping with Jev right now.