Kakao said on Nov. 6 that it had released its multimodal decision-making model "Kanana-2-v-decide" on Hugging Face. The model is designed to let AI agents make the judgments they need quickly while carrying out tasks, returning results in the form of probabilities, choices or scores rather than generating sentences as a typical large language model does, allowing AI agents to move on to the next step faster. It carries the Kanana Open License, which permits commercial use as well as research. Kakao developed the model on the basis of Kanana-2-v-3b, its own multimodal language model, and because it can analyze text and images at the same time, it can be used to classify customer inquiries, select tools for AI agents and identify product images. According to Kakao, Kanana-2-v-decide recorded an average response time within 0.04 seconds on a single Nvidia H200 graphics processing unit, which by Kakao's own measurements is about 43 times faster than GPT-6 Luna. In performance evaluations, the model posted the highest scores among the models compared on Kanana Decision Bench, Kakao's own evaluation data set of 15,909 Korean-language questions, and on Decision Index, a publicly available evaluation data set. Kakao plans to apply the model first to internal projects such as web search and automatic classification of product categories.