JEV-27B-Uncensored-GGUF
Static GGUF quantizations by mradermacher of SargeDev's JEV-27B-Uncensored, a Jev 1.13-distilled decision model.
Overview
This Hugging Face repository holds static GGUF quantizations, made by mradermacher, of SargeDev/JEV-27B-Uncensored. The base is a Qwen3.8 27B decision model distilled from Jev 1.13 on the jev-distill-corpus-v3 dataset, with an abliterated (uncensored) variant. Tags indicate support for noul, choice and score typed decisions with calibrated probabilities.
It is for people who want to run that model locally with llama.cpp-compatible tools. Files range from Q2_K at 10.8 GB upward, with Q4_K_S marked fast and recommended, plus small multimodal projector files. The card contains no description of the model itself beyond tags, and the decision readout requires tooling that understands the model's decision format; plain GGUF chat will not produce typed probabilities on its own. Users should check the upstream SargeDev model card for intended use.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Same typed noul/choice/score decisions as the upstream model(inferred) | — |
How it works
- Pick a quantization by size and quality
- Download the GGUF (and mmproj if needed)
- Run with llama.cpp-compatible tooling that supports the decision readout
Strengths
- Range of sizes from about 11 GB
- Apache-2.0 licence
Caveats
- No upstream description; behaviour must be inferred from the base model
- Uncensored/abliterated variant removes refusal behaviour
- Low-bit quants reduce quality
At a glance
- Type
- Model
- Runtime
- GGUF / llama.cpp
- Has code
- no
- License
- Apache-2.0
Use cases and integrations
How curated lists describe it
No upstream description was published.
wh000wh000/awesome-jev-live
Frequently asked questions
What is this?
GGUF quantizations of SargeDev/JEV-27B-Uncensored for local use.
Which quant should I use?
The card marks Q4_K_S as fast and recommended.
Will Ollama chat give me decisions?
Not on its own; typed decisions need a readout that understands the model's format.
Related projects
Mangaba AI's local, calibrated typed-decision model for Portuguese text, trained to optimize the Brier score; proprietary since v5.
ProjectproductionPython- blink★ 22
Open one-pass typed-decision model with a dependency-free C99 runtime that also builds to a 66 KB WebAssembly module.
ServicebetaC - Foq★ 11
Local open alternative to Jev using a 1.58-bit ternary 8B model to answer booleans, choices, scores and Pydantic extraction in ~25 ms.
ProjectbetaPython Runs the open Laya decision model entirely in a browser tab with WebGPU and ONNX Runtime Web: no server, no API key.
ServicebetaTypeScriptllama.cpp-based C++ server that turns any LLM into a System One decision model, reading answers from next-token logits.
ProjectbetaC++Three open 4B decision models for English, Turkish and German that serve the Jev /v1/systemone API on a CPU in about 3 GB of RAM.
ServicebetaPython