A New Chapter in Reasoning
In January, we introduced SAGE-32B as our most capable model for agentic tasks. Today, we are rebranding this architecture as SAGE 2.4 Actus. This isn't merely a name change—it represents our commitment to the Actus lineage to spearhead our future research into truly autonomous, long-horizon planning models.
Under the hood, SAGE 2.4 Actus leverages the identical robust 32-billion parameter foundation developed for SAGE-32B. Enriched through rigorous Iterative Distillation and inverse reasoning, it excels at tool calling, error recovery, and complex step-by-step problem solving.
Uncompromising Performance
Fig 1: SAGE 2.4 Actus outperforming comparable open-weights models on key reasoning tasks.
Agentic ReasoningActus Rebrand32B Parameters2026
Authors
SAGEA AI Research