Jev directory
CommunityJev Models & Reimplementations
Decision-1.0-Nox-4B
Nox-4B, the 4B model in vLLM Semantic Router's Decision 1.0 family: give it a state, questions and possible answers, and it returns typed decisions and probabilities with labels defined at runtime. The card reports 73.09% weighted accuracy across 3,766 decisions.
- Reading time
- 2 min
- Published by
- Community
- Author
- vLLM Semantic Router
- Last verified
- Oct 2026
Highlights
- 4B decision model on Qwen/Qwen3.5-4B serving choice, noul and score questions with labels defined at runtime.
- Card-reported 73.09% weighted accuracy over 3,766 decisions and 54 tasks, 2.99 points above Kev-4B.
- Choice routes among 2-255 actions, noul returns P(true) and score uses 2-10 ordered rubric descriptions.
- A complete question, candidates and state are limited to 16,384 tokens; longer questions fail with max_length_exceeded.
- Runs with stock Transformers through trust_remote_code; Apache-2.0, and serving lives in vLLM Semantic Router.
Quickstart
bash
pip install "transformers>=5.17" torch safetensors huggingface_hub
hf download vllm-sr/Decision-1.0-Nox-4B --local-dir Decision-1.0-Nox-4BWatch out
Card reviewed 2026-10-17; the 73.09% weighted accuracy and the 2.99-point gain over Kev-4B are self-reported, and the card discloses that the overall weights were chosen after observing results. Latency was measured on one AMD gfx942 GPU, and this release bundles no serving code.
Questions about this entry
Short answers drawn from the linked project or page.
Nox-4B, a 4B decision model from vLLM Semantic Router's Decision 1.0 family. Give it a state, questions and possible answers, and it returns typed decisions and probabilities with the labels defined at runtime. Nox is Latin for night, says the card.
Its system_one method takes and returns the same System One request and response bodies as the Decision runtime, and nothing is generated. The card's comparison table lists Jev at 81.05 overall beside Nox-4B's 73.09 as an external reference.
Choice for routing among 2 to 255 actions, returning the selected ID plus a distribution; noul for checking a condition against evidence, returning P(true); and score over 2 to 10 ordered rubric descriptions, returning an expected index plus a distribution.
With stock Transformers and trust_remote_code=True: AutoModel.from_pretrained("vllm-sr/Decision-1.0-Nox-4B", trust_remote_code=True), then model.system_one(state=..., questions=...). pipeline("decision", model=...) accepts the same request body, and Apache-2.0 applies.
73.09% weighted accuracy across 3,766 decisions and 54 tasks, 2.99 points over Kev-4B on the same benchmark, with area scores such as Decisions 83.00, Composition 51.79 and Inference 86.25. The card chose these overall weights after seeing results.
A complete question, its candidates and the state are capped at 16,384 tokens; a longer question fails the whole request with max_length_exceeded and nothing is truncated. On a GPU the backbone runs BF16 with an FP32 head, on CPU everything is FP32.
The vLLM Semantic Router Decision runtime loads and serves Choice, Noul and Score requests; this release bundles model data and Transformers loading code only. The weights are accelerator-agnostic because which hardware can serve them is decided by the runtime.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Both Jev 1.13 and 1.14 score 29 of 30 on 30 Turkish buyer and seller messages
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 NewsMore Jev resources like this
The v2.2 PEFT LoRA adapter for Standard Thinking's Standard One 3B decision model, shipped with a merge recipe for Ministral-3-3B-Instruct; serving requires the merged checkpoint and the shared server code in the StandardOne-8B repository.
Jev Models & Reimplementations#community#open-models#local
Community
Standard Thinking's 8B LoRA adapter for the Standard One decision model: merge it with the supplied recipe and serve it, and it scores choice, noul and score labels over a scenario with calibrated probabilities via POST /v1/systemone, never generating text.
Jev Models & Reimplementations#community#open-models#local
Community
The smallest decider from Mapika: a 1.4 GB bf16 Qwen3.5-0.8B-Base fine-tune that returns calibrated probabilities for typed Choice, Score and Noul questions in one forward pass and runs 1.5x faster than decider-2b, within one to four points on routing and lookups.
Jev Models & Reimplementations#community#open-models#local
Community
A local daemon that pulls open decision models and serves them on TypeSafe's /v1/systemone, so a Jev client switches hosts with TYPESAFE_BASE_URL.
Jev Models & Reimplementations#community#rust#open-models
Community
ollaya-dev/ollayaRust
ollaya-dev/ollayaRustA local server that reads next-token probabilities from a fine-tuned Spark model and returns Choice, Score, and Noul answers with zero generated tokens.
Jev Models & Reimplementations#community#python#open-models
Community
Rizzo-AI-Academy/rizzo-flowPython
Rizzo-AI-Academy/rizzo-flowPythonA local /v1/systemone server that reads label probabilities from a GGUF chat model, with prompt repetition and a confidence shrink toward uniform.
Jev Models & Reimplementations#community#python#open-models
Community
Mushroom-Systems/lichenPython
Mushroom-Systems/lichenPythonFrom 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.