Artemis Edge decision models
TurinTech GGUF builds of Clef, Clef-flash, pplx-decider and Kev for stock llama.cpp at /v1/systemone, at about 27-30% of full size.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
Artemis Edge is a Hugging Face collection from TurinTech with independent GGUF quantizations of several open decision models: Clef, Clef-flash, pplx-decider-v1.1-27b and Kev (4B and 9B).
Per the listing, the default files are 27 to 30% of the original size and give the same answer as the BF16 original on 92.2 to 96.8% of 2,000 typed-decision test cases in TurinTech's runs, with each card reporting held-out decisions and comparisons with public quantizations. They are Apache-2.0 like the originals, served by stock llama.cpp at /v1/systemone, and as of October 13, 2026 described as the only public pplx-decider GGUFs that run on stock llama.cpp.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Other | Does the quantized model give the same typed decision as the original?(inferred) | — |
How it works
- Download a GGUF from the collection
- Serve with stock llama.cpp
- Call /v1/systemone with Jev-style requests
Strengths
- Much smaller files for edge hardware
- Agreement with originals reported per card
- Stock llama.cpp, no custom runtime
Caveats
- Agreement figures are TurinTech's own runs
- Quantization changes some answers
Getting started
Download a model from the Hugging Face collection and serve it with llama.cpp.At a glance
- Type
- Model
- Status
- production
- Runtime
- llama.cpp (GGUF)
- Has code
- no
- License
- Apache-2.0
Use cases and integrations
How curated lists describe it
TurinTech's independent GGUF builds of Clef, Clef-flash, pplx-decider-v1.1-27b, and Kev (4B and 9B). The default files are 27 to 30% of the original size and give the same answer as the BF16 original on 92.2 to 96.8% of 2,000 typed-decisions test decisions in TurinTech's runs; each card also reports held-out decisions and comparisons with public quantizations. Apache-2.0 like the originals and served by stock llama.cpp at `/v1/systemone`; as of October 13, 2026, these are the only public pplx-decider GGUFs that run on stock llama.cpp.
AnotiaWang/awesome-decision-models
Frequently asked questions
Which models are included?
Clef, Clef-flash, pplx-decider-v1.1-27b and Kev 4B and 9B.
How close are they to the originals?
Same answer on 92.2-96.8% of 2,000 test decisions in TurinTech's runs.
What runs them?
Stock llama.cpp at /v1/systemone.
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