Nace AI has released Drex 2.0 for macOS, a version designed to run entirely on the user's machine so that AI agents can make decisions without sending any data to the cloud.
Nace AI · Drex 2.0 for macOS
A local model that lets AI agents choose without sending data to the cloud
Drex 2.0 runs entirely on the user's machine. It scores multiple-choice options so an agent can rank and pick one, keeping the data behind that decision on the device.
How a decision works
Drex scores candidate answers rather than generating free-form text.
Step 1
Agent poses a question
Options are fixed and framed as multiple choice.
→
Step 2 · On device
Drex 2.0 scores each option
Likelihood is computed locally. No data leaves the Mac.
→
Step 3
Ranked basis for selection
The agent receives an ordered list and picks one option.
Why it matters
Sensitive files, internal records and personal information can stay on the device during the decision step.
Reduces dependence on remote cloud models for agent workflows.
Open-weight format means developers can, in principle, inspect and adapt the model.
What is still unclear
No independent benchmarks or accuracy figures have been published.
Hardware requirements and licensing terms were not detailed.
Multiple-choice framing may not fit many real agent tasks beyond fixed options.
What to watch
Three questions will determine whether local decision models matter
01
How does Drex 2.0 perform on independent tests?
02
How easily does it integrate with existing macOS agent frameworks?
03
Will other vendors adopt similar local decision models?
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...