Fastino Labs Launches GLiDE to Outperform Top AI Decision Models

GLiDE, or Generalized Lightweight Decision Engine, secures its lead by outperforming existing benchmarks across five key categories. While typical models rely on fixed, single-pass scoring, GLiDE utilizes a dual-path architecture. It resolves routine tasks via a fast, initial pass, but automatically triggers deeper reasoning cycles when confidence remains low. This approach allows the system to balance computational speed with the precision required for difficult causal and mathematical reasoning.

In testing against the Decision Index 0.2.1 suite, the model achieved significant gains in Knowledge and Reasoning, Tools and Automation, and human-centric tasks. Most notably, it reached 92.6% accuracy on CRUXEval for code reasoning and 88.7% on CLadder for causal analysis. CEO George Hurn-Maloney suggests this functionality creates a dedicated layer within the AI stack, reducing the reliance on costly, general-purpose frontier models for routine agentic navigation.

The system outputs structured data—providing a selected action, a confidence score, and probability distributions for all options—which allows developers to integrate it directly into existing production pipelines. With a 40,000-token context window, GLiDE is designed to ingest full policy documents and operational runbooks, enabling it to manage high-risk tasks such as financial approvals, incident diagnostics, and complex contract verification. The model is accessible immediately through the Fastino API.