CommunityJev Models & Reimplementations
Jev-Style 0.8B Decision v3 GGUF
GGUF builds of the Jev-Style v3 decision model on Qwen3.5-0.8B: a 0.53 GB Q4_K_M file the card says is decision-identical to FP32 on 240/240 parity rows, served through a pinned llama.cpp scorer.
- Published by
- Community
- Author
- chaoliangUNSW
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRuns locally
Highlights
- Q4_K_M is 0.53 GB and matched PyTorch FP32 on 240 of 240 parity rows and 6 of 6 long prompts.
- The GGUF path needs a custom jev-score binary built against a pinned llama.cpp commit; llama.cpp's own server will not emit decisions.
- The 0.8B model scores 79.2% on 2,000 typed decisions, 2.6 points over Laya's fine-tuned checkpoint in the card's comparison.
- Reads up to 25,600 tokens of input inside a 32,768-token runtime window, with no truncation — over-budget inputs raise an error.
Watch out
Some training data carries restrictive or unclear terms and some rows are outputs of OpenAI and Anthropic models; the training pipeline is not included. The project is independent of TypeSafe, Jev, Laya, and Qwen, and long option lists are scored in chunks through one softmax.
Questions about this entry
Short answers drawn from the linked project or page.
Q4_K_M at 0.53 GB is the practical default: the card reports it decision-identical to FP32 on 240 of 240 parity rows and roughly 2.4 times smaller than the 2B v2's Q4_K_M. F16 is the runtime default and the basis for the card's latency figures.
Not for decisions. The repo builds a one-time jev-score binary against a pinned llama.cpp commit (441df11f65ea0b6d0c72965aaf70c8241070ddcb or later); the GitHub tooling serves a systemone-compatible API, and plain chat completion does not produce decisions.
The card reports 79.2% on 2,000 typed decisions, 2.6 percentage points above Laya's fine-tuned checkpoint in its own comparison. That comparison is the card's own rather than a third-party rerun.
Up to 25,600 tokens of input inside a 32,768-token runtime window with no truncation. Inputs beyond that budget raise an error instead of being silently shortened, so an over-long request fails rather than returning a quietly clipped answer.
The project says some training data carries restrictive or unclear terms and some rows are outputs of OpenAI and Anthropic models, and the training pipeline is not included. It states it is independent of TypeSafe, Jev, Laya and Qwen.
Long option lists are scored in chunks that are combined through a single softmax, so the option count is not limited by a fixed per-call window the way a single-shot call would be. The card does not publish a maximum option count.
It is built on Qwen3.5-0.8B, with Q4_K_M at 0.53 GB and F16 as the runtime default. The v3 GGUF is roughly 2.4 times smaller than the 2B v2's Q4_K_M quant.
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
An open distillation of Jev 1.13 onto Qwen3.8-27B: a 108.9M-parameter decision block answers typed questions in one prefill pass and adds a POST /v1/decide route beside a full System 2.
Jev Models & Reimplementations#community#open-models#local
Community
The 9B member of AutoTrust's Jev students: a 40.2M-parameter decision block on Qwen3.5-9B with 0.019 mean KL to Jev 1.13 and 70.7% HumanEval left untouched, small enough for one modest GPU.
Jev Models & Reimplementations#community#open-models#local
Community
AutoTrust's Gemma-based decision model with adaptive thinking: System 1 answers below 0.8 confidence hand off to the full Gemma 4 26B and mix the two by half, behind POST /v1/decide.
Jev Models & Reimplementations#community#open-models#local
Community
A 400M-parameter Jev-compatible decision model with a browser build: 0.796 on typed decisions against Laya's 0.766, 17 ms per question on an L40S, and an ONNX file that runs in a tab.
Jev Models & Reimplementations#community#open-models#local
Community
1.3kGitHub stars
A local yes/no decision model that speaks Jev's wire format on a laptop CPU, refusing Choice and Score until those question types ship.
Jev Models & Reimplementations#community#python#open-models
Community
feder-cr/jevC++
feder-cr/jevC++8GitHub stars
A Python server in front of llama-server that reads one-token label probabilities and exposes them as POST /v1/systemone.
Jev Models & Reimplementations#community#python#open-models
Community
khimaros/verdictPython
khimaros/verdictPythonFrom the Jev guides
Original write-ups on what this entry is for.
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
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
From the Jev community
Posts from builders shipping with Jev right now.