A recent study evaluates the performance of large language models in system dynamics AI assistance, comparing cloud-based proprietary APIs with locally-hosted open-source models. The evaluation is based on two benchmarks: the CLD Leaderboard, which assesses the models' ability to extract causal loop diagrams, and the Discussion Leaderboard, which tests their capacity for interactive model discussion and feedback explanation. The study reveals significant differences in performance between cloud and local models, with some open-source models outperforming their proprietary counterparts in specific tasks1. The findings have implications for the development and deployment of system dynamics AI assistants, highlighting the need for careful consideration of model architecture and hosting options. This research matters to practitioners because it informs decisions about the trade-offs between cloud-based and local AI solutions, affecting the accuracy, efficiency, and security of system dynamics modeling and analysis.