The final 32 teams for the upcoming World Cup have been confirmed, and the results of a prediction test involving 12 major AI models have been released.

Among the participants, Tencent's Hunyuan model took the lead by correctly predicting 29 out of the 32 qualifiers. The models from MiniMax and Iflytek Co.,Ltd.'s Spark each accurately forecasted 28 teams. Other models, including those from DeepSeek, Zhipu AI, and Lenovo's Tianxi AI, all achieved a score of 27 correct predictions.

The analysis indicates a high level of consensus among the AI models when predicting traditional powerhouses and widely expected qualifiers. The key differences in performance emerged in forecasting the qualification of more marginal teams, such as Bosnia and Herzegovina, Cape Verde, DR Congo, and Ghana, as well as in predicting the elimination of nations like Iran, South Korea, and Uruguay.

These results demonstrate that AI models can now maintain a high accuracy rate in large-sample predictions. However, the discrepancies between models become more pronounced when dealing with borderline qualification spots, groups with potential upsets, and specific match outcomes like wins, draws, or losses. The exercise shows that accurate forecasting is not merely about understanding football, but also tests a model's comprehensive ability to process real-time information, tournament rules, team form, and inherent uncertainties.