NaceAI launches Drex sub-6B model for agent decisions
The implication is that generation is wasted compute when the required output is a choice among known options, and a small classifier-style model can serve that role at a fraction of the cost. Teams building agent routing or tool-selection layers can swap a generative model for Drex in those steps while keeping a larger model for prose tasks. Compared with typical 100B-plus LLMs used for routing, the sub-6B size points to lower latency and cost.
NaceAI has launched Drex, a sub-6B decision model that outputs option probabilities instead of generated prose.
In a single pass, it takes facts as input and returns a probability for each available option.This removes the need to generate a reasoning paragraph and then parse it back into a tool call or routing decision.
NaceAI positions Drex for agent routing, tool selection, reranking and policy checks, where systems must choose from a known set of options.
Most agent stacks use a generative model to make decisions that never needed prose in the first place.
@NaceAI just launched Drex, a sub-6B decision model that outputs option probabilities instead of generating prose.
facts in, a probability for each available option out, in one pass. No reasoning paragraph to generate, then parse back into a tool call or routing decision.
And that makes a huge difference for agent routing, tool selection, reranking, policy checks, and other places where the system ultimately needs to choose from a known set. And Nace explicitly positions Drex around those workloads.
If the output you need is a decision, generation can be unnecessary compute.
This report is categorized under AITOP's "Models" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include NaceAI, Drex, agents, small models.
Source attribution: curated from rohanpaul_ai (@rohanpaul_ai).
