Nace.AI’s Drex Max Ranks No. 1 Among Decision Models on Artificial Analysis’s Decision Index
Drex helps enterprises and AI agents make decisions with probabilities that guide when to act and when to request
Press Release Disclaimer: This is a press release distributed through the XPR Media network. It has not been independently verified by our newsroom.
Nace.AI today announced that Drex Max v1.1 ranks No. 1 among decision models on the Artificial Analysis Decision Index, an independent benchmark of decision quality.
This press release features multimedia. View the full release here: https://www.businesswire.com/news/home/20261113545014/en/
Drex Max v1.1 leads the decision models displayed on the Artificial Analysis Decision Index; chart captured November 11, 2026.
Large language models made today’s AI agents possible. In production, those agents can make thousands of small decisions a day: which tool to call, how to classify a document, when to involve a person. Decision models are built for these repeated choices, returning probabilities that software can use alongside business rules to act or escalate.
In an invoice-review workflow, for example, a developer gives Drex the invoice, its purchase order and the company’s approval rules, and asks it to choose among “ready for approval,” “missing evidence” and “policy exception.” Drex returns a probability for each. The workflow applies the customer’s thresholds and approval controls to decide what proceeds and what goes to a reviewer.
The Artificial Analysis Decision Index combines Decision Scores across five evaluations with a separate Symmetry Score. It rewards correct answers and well-calibrated confidence, penalizes confidently wrong answers, and assesses consistency when questions are rephrased. These measures matter when a confidently wrong answer could let an invoice that should be held keep moving.
“LLMs design the workflow. Nace small models run it. Drex is the one that decides,” Dos Baha, Founder and CEO, Nace.AI
Drex Max v1.1 supports text and image inputs and a 262K-token context window, enabling decisions informed by documents, images and extended context. Artificial Analysis reports 283 milliseconds for its “Time per task” metric.
Benchmark note: The No. 1 ranking among decision models was verified on Artificial Analysis’s model page on November 12, 2026. The November 11 chart displays 18 of 19 models; Drex Max v1.1, SubQ Snap 1.2 and GLiDE v1.2 each show a rounded score of 73. See the independent benchmark results.
Developers can get started with Drex at nace.ai/developers.
About Nace.AI
Nace.AI builds specialized small models that run enterprise workflows. Headquartered in Palo Alto, California, the company is backed by Walden Catalyst and General Catalyst. Learn more at Nace.ai
View source version on businesswire.com: https://www.businesswire.com/news/home/20261113545014/en/
Media gallery
