Decentralized AI platform FLock.io has open-sourced its decision model, this-that-model-1.0, trained on its own framework. The team said a 68-question decision test showed 94.1% accuracy, above GPT-5.6 at 92.6%, DeepSeek V4.1 Flash at 79.4%, Kimi K3 at 77.9%, and Jev at 76.5%.
According to Foresight News, the model can run locally and return decisions and probabilities in a single forward pass. The published local latency on an RTX 5080 Laptop GPU is about 30.9 milliseconds, which the team said is faster than other models released at the same time.
FLock said the open-source release gives its agents and business workflows independently deployable decision-making capabilities and reduces reliance on centralized cloud APIs. The company, which is a United Nations Development Programme AI partner, has deployed decentralized AI and solutions in several United Nations member states, including Malaysia and the Dominican Republic.