01What Is DREX 2.0 and How Does It Work?

DREX 2.0 is a decision model from NACE.AI that was released on October 6th, 2026. Unlike most AI models that generate long text answers, DREX 2.0 makes fast, structured decisions. It returns probabilities for the options you define instead of writing paragraphs.

You give it context, a typed question, and a list of allowed answers. It scores each option. Your system then takes action based on the highest score.

No fluff, no extra text. Just a clean decision.

DREX 2.0 runs on a mixture of experts architecture. It has about 35 billion total parameters but only activates around 3 billion per token. Think of it like a team of specialists.

You only pull in the experts you need for each decision. That keeps it fast without losing quality.

02Key Features of DREX 2.0

DREX 2.0 was built on top of QEM 3.6-35B-A3B. NACE added a decision scoring layer on top. Instead of generating open-ended text, it returns probabilities for the options you define.

  • Decision scoring layer: Gives you scores for each option, not text.
  • Speed improvement: NACE says DREX 2.0 is up to 38% faster than version 1.5 on inputs between 32,000 and 128,000 tokens.
  • Better decision quality: On the public decision index 0.3.1, DREX 2.0 scores 61.25, up from 58.08 in version 1.5. That is a 3.17 point improvement across 37 public benchmarks.
  • Email classification: DREX 2.0 scores an AUC of 0.984 on email classification. Version 1.5 scored 0.972.
  • Side bench accuracy: It hits 96.3% zero shot accuracy and 96.5% with a few examples given up front.

03How to Set Up DREX 2.0

Getting started with DREX 2.0 is straightforward. Here are the steps you can follow based on the 30-day roadmap inside the AI Profit Boardroom.

  1. Set up your NACE API access. You need an API key to use DREX 2.0.
  2. Define your first decision schema. This includes the context, the question, and the allowed answers.
  3. Connect DREX 2.0 to your lead intake system. For example, a form or email system.
  4. Run your first live routing test. Use real data to see how the model scores options.
  5. Automate the action. Have your system pick the top score and trigger the appropriate action without human involvement.

The AI Profit Boardroom provides daily tutorials and coaching calls to help you build this stack. You can show up live with your specific setup and get help from people who are already running decision models in production.

04How to Use DREX 2.0 in Your Business

Here is a practical example from the episode. Imagine someone signs up for your service. Their information hits your system.

You have three possible actions: invite them to a coaching call, send them an onboarding sequence, or tag them for a follow-up question.

With DREX 2.0, you feed it the sign-up context. You ask it which of those three actions fits best. You get back three scores.

Your system picks the top one and fires automatically. No human needs to touch it.

Another example is sorting community questions before a coaching call. You build a DREX layer that reads submitted questions and scores each one across three options: answer live, add to async queue, or redirect to an existing tutorial. The coach walks in and the high-priority questions are already at the top.

That saves prep time every week.

05Results and Benchmarks from the Episode

NACE claims DREX 2.0 is up to 38% faster than version 1.5 on inputs between 32,000 and 128,000 tokens. The actual speed depends on your setup, hardware, and how you use the API. But even as a vendor claim, it is meaningful for systems making hundreds of decisions a day.

On the public decision index 0.3.1, DREX 2.0 scores 61.25. Version 1.5 scored 58.08. That is a 3.17 point improvement across 37 public benchmarks.

This measures decision quality, not speed.

For email classification, DREX 2.0 scores an AUC of 0.984. Version 1.5 scored 0.972. JEV 1.13.0 scored 0.981.

So DREX 2.0 tops that leaderboard. On the side bench, it hits 96.3% zero shot accuracy and 96.5% with a few examples. JEV scores 98.0% on the same benchmark.

DREX 2.0 wins some comparisons but not every single one.

06Tips, Limits, and Mistakes to Avoid

Most business owners still route everything through a general purpose model. That costs more per token, gives slower outputs, and misses the dedicated decision layer. If you want an operational edge, start using decision models now.

  • Define clear options: The model works best when your allowed answers are specific and distinct.
  • Test with real data: Run a live routing test before going fully automated. This helps you catch issues early.
  • Don't ignore the limits: DREX 2.0 does not win every benchmark. For example, JEV scores higher on the side bench (98.0% vs 96.3%).
  • Watch your input size: The speed claim applies to inputs between 32,000 and 128,000 tokens. Performance may vary outside that range.

07Who Should Use DREX 2.0 and Why It Matters

DREX 2.0 is for anyone using AI automation in their business. If you run lead qualification systems, content routing, community management automation, or email classification, this model can help you make faster and cleaner decisions.

It matters because it saves time and money. You pay less per token compared to general purpose models. Your automation runs faster.

And the decisions are structured, so your system can act on them immediately.

The businesses building with decision models now are going to have a real operational edge. It compounds fast as they scale. The AI Profit Boardroom offers a 30-day roadmap, daily tutorials, and coaching calls to help you get started.

You can join the free AI Success Lab for access to prompt templates and community support.