[LM5.2 Ranks Second in Long-Term Commercial Simulation Evaluation, Kimi and Minimax New Versions Show Divergence] Golden Finance reports that the latest Vending-Bench 2 evaluation released by Andon Labs shows that the open-source model GLM 5.2 successfully secured second place. The evaluation simulated the virtual operation of a vending machine enterprise over a 365-day period, inputting daily inventory and financial data into the model, which then made decisions on restocking, pricing, and other actions via API calls. The goal was to assess the decision-making consistency of large language models in long-term tasks. Data analysis revealed that GLM's successive versions demonstrated a strong linear growth trend in performance during the evaluation, with average monthly profit increasing by nearly $1,000 (GLM 5 had an average score of $4,432, while GLM 5.1 improved to $5,634). Compared to GLM's steady progress, other mainstream domestic models showed varied performance in their latest versions. Kimi K2.7 Code performed worse in the evaluation compared to its predecessor, Kimi K2.6. Minimax M3 achieved significant improvement over its predecessor, M2.5, but its overall profitability still lagged far behind the Kimi and GLM series models.