Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

Drex decision models

Nace.AI's open Drex decision models (v2.0 MoE, v1.5 9B, diffusion DLM) serving /v1/systemone locally via Python, llama.cpp or Ollama.

ServicebetaNace.AI Open RAIL-M
View on GitHub ↗

Overview

Drex is a family of open decision models from Nace.AI. You pass a state and named typed questions (yes/no, multiple choice or rating scale) and get a probability for every option; nothing is generated. Every model serves the same POST /v1/systemone API as the hosted Drex API and runs locally from a Python server or from Nace.AI branches of llama.cpp and Ollama.

The lineup includes Drex v2.0, a 35B mixture-of-experts model with about 3B active parameters that the README reports at 61.25 on the public Decision Index 0.3.1; Drex v1.5 on a distilled Qwen 3.5 9B backbone, reported at 58.08 and as the top model under 10B; and Drex DLM on NVIDIA's Efficient-DLM-8B diffusion backbone, with a Q8_0 GGUF for Apple silicon and CUDA. Licences differ per model: v2.0 and v1.5 use a Nace.AI Open RAIL-M licence, and Drex DLM weights are CC-BY-NC-4.0. An agent skill is also available.

What the decision model decides

TypeQuestionAnswers
ChoiceNamed typed questions (yes/no, multiple choice, rating scale) over a state—

How it works

  1. Download weights or a GGUF
  2. Serve with the Python server, llama.cpp or Ollama branch
  3. POST state and questions to /v1/systemone
  4. Read per-option probabilities

Strengths

  • Several sizes and architectures
  • Same API locally and hosted
  • Runs on llama.cpp and Ollama

Caveats

  • Licences vary; DLM weights are non-commercial
  • Requires forked llama.cpp/Ollama branches

Getting started

See the model pages under models/ and the Hugging Face weights (nace-ai/drex-v1.5, nace-ai/drex-v2.0).

At a glance

Type
Service
Status
beta
Language
Python
Runtime
Python server, llama.cpp, Ollama
Has code
yes
License
Nace.AI Open RAIL-M

Use cases and integrations

How curated lists describe it

  • Nace.AI's open decision models, Drex v1.5 on a 9B Qwen-based backbone and Drex DLM on NVIDIA's Efficient-DLM-8B diffusion backbone, each answering `POST /v1/systemone` from a Python server or forks of llama.cpp and Ollama. Formerly `drex-dlm`; Drex DLM weights under CC-BY-NC-4.0.

    MrJev/awesome-jev

Frequently asked questions

Which Drex model should I use?

v1.5 is the README's pick under 10B; v2.0 scores higher but is a 35B MoE.

Is it Jev-compatible?

It serves the same /v1/systemone API shape.

Can I use them commercially?

Check each model's licence; Drex DLM weights are CC-BY-NC-4.0.

Related projects

  • brier

    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
  • browser-laya

    Runs the open Laya decision model entirely in a browser tab with WebGPU and ONNX Runtime Web: no server, no API key.

    ServicebetaTypeScript
  • ifReflex.cpp

    llama.cpp-based C++ server that turns any LLM into a System One decision model, reading answers from next-token logits.

    ProjectbetaC++
  • jevalt

    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