Drex Max v1.1 Tops Decision Index

Drex Max v1.1 is compelling because it optimizes for the operational bottleneck in agents: making reliable choices quickly, not generating more text.

• –Returns probabilities directly, making confidence thresholds and human escalation easier to implement
• –Multimodal inputs extend it beyond text routing into screenshots, documents, and visual agent workflows
• –Its benchmark lead is paired with substantially lower cost and latency than frontier general-purpose models
• –The result is especially relevant for research triage, classification, tool selection, and browser automation
• –Open weights could make the model far more useful for privacy-sensitive and local deployments