Fastino Labs unveils GLiDE, the Generalized Lightweight Decision Engine, touted as the first decision model with adaptive reasoning and a solid lead on the Decision Index 0.2.1 with a score of 64.81 versus Jev at 57.91.

GLiDE features a 40,000-token context window, enabling handling of full policies, contracts, runbooks, and support histories.

Its output is structured: one chosen action, a confidence score, and a probability for every option, allowing immediate action, threshold setting, or human review without free-form parsing.

GLiDE uses adaptive thinking, with a fast initial pass resolving about two-thirds of tasks and deeper reasoning engaged for the remaining one-third.

In hard-decision testing, GLiDE shows strong gains in Knowledge and Reasoning and Tools and Automation, plus higher code reasoning accuracy (CRUXEval: 92.6% vs 73.0%) and causal reasoning (CLadder: 88.7% vs 72.6%). Specifically, 62.9 vs 51.4 in knowledge/reasoning and 83.5 vs 75.1 in tools/automation.

GLiDE outperforms Jev across all five evaluation areas and 31 of 38 benchmarks, and is now available via the Fastino API.

CEO and co-founder George Hurn-Maloney notes decision-making will become a separate AI stack layer, highlighting GLiNER’s 50+ million downloads and positioning GLiDE as Fastino’s best decision model to date.