CatalystTech/AI
Ant's Jev-style model is here! The 27B version ranks first in overall score, saving 90% of tokens.
Ant's 27B model ranks first overall while saving 90% on tokens.
Context
What happened: Ant Digital Technologies released LingDT-Decision-1.0, a decision model in 4B and 27B versions built for high-frequency tasks in customer service, risk control and finance, with native multi-select and ordered (DAG) decision features.
Why it matters: Ant says the 27B version's overall score exceeded Microsoft-Decision-1, Jev 1.13.0 and Fastino/GLiDE, and both versions led 16 decision models on multi-select tasks. The release moves decision models into enterprise workflows after Jev's rise drew Microsoft and Perplexity into the category.
What changed
Ant Digital Technologies released LingDT-Decision-1.0, a decision model in 4B and 27B versions built for high-frequency tasks in customer service, risk control and finance, with native multi-select and ordered (DAG) decision features.
Why it matters
Ant says the 27B version's overall score exceeded Microsoft-Decision-1, Jev 1.13.0 and Fastino/GLiDE, and both versions led 16 decision models on multi-select tasks. The release moves decision models into enterprise workflows after Jev's rise drew Microsoft and Perplexity into the category.