FLock.io Launches Edge‑Optimized Decision Model to Replace Slow, Costly Cloud LLMs

As enterprises accelerate AI adoption, reliance on centralized cloud‑hosted LLMs has created major privacy, regulatory, and latency risks. FLock.io has introduced THIS / THAT Model, an open‑source, lightweight decision model built on Alibaba Cloud’s Qwen3.5‑2B architecture, designed to run locally on edge devices and eliminate the need to send sensitive data through external cloud APIs.

The launch comes amid growing interest in “System One” decision models—fast, deterministic systems built for predefined answer sets such as yes/no. Unlike traditional LLMs that generate open‑ended text, decision models like TypeSafe’s Jev and FLock.io’s THIS / THAT are optimized for high‑volume classification tasks where natural‑language generation is unnecessary and costly.

FLock.io CEO Jiahao Sun said the industry is hitting a performance ceiling: “For many critical enterprise tasks, generating open‑ended text is an unnecessary cost and bottleneck. With THIS / THAT Model, we are demonstrating that specialized 2B‑parameter models running locally can outperform massive cloud LLMs in speed, accuracy, and cost‑efficiency—while keeping data strictly on‑device and reducing compliance risk.”

THIS / THAT delivers local inference in approximately 31.4 milliseconds—roughly 30x to 100x faster than cloud‑hosted LLMs—and operates fully offline with zero dependency on external servers. It avoids network latency, prevents cloud data collection, and maintains functionality during outages. Despite having only 1.88B parameters (<0.04% of Claude Opus 5), the model achieved an 87.8% success rate across 1,710 complex decision‑making tasks and outperformed Jev in local‑geometry benchmarks. Three weeks after launch, it has already processed 2.5 million decisions.

The model is built on Qwen3.5‑2B, a compact foundation model that runs on laptops and smartphones. It was trained on synthetic data generated using Qwen Max and incorporates findings from FLock.io’s ACL‑accepted research papers, GraphSynth and CircuitSynth. THIS / THAT is designed for edge environments where ultra‑low latency, privacy preservation, and predictable outputs are essential.

FLock.io and Alibaba Cloud partnered in June 2025, marking the first deep integration between a major Web2 LLM provider and a decentralized training platform. FLock.io specializes in federated learning and privacy‑preserving AI infrastructure, enabling enterprises to train and deploy custom models locally while maintaining full data ownership and regulatory compliance.