The article by Thomas Reichler and Junsu Kim evaluates the performance of coupled climate models in simulating today's climate. Coupled climate models are sophisticated tools used to understand and predict climate change, but they are not perfect due to incomplete theoretical understanding and simplifying assumptions. The study aims to objectively quantify the agreement between model and observations using a single performance index derived from a broad group of variables. This approach is novel compared to previous model intercomparison studies, which often focused on specific processes or considered a narrow range of models. The study includes model output from three generations of climate models: CMIP-1, CMIP-2, and CMIP-3 (IPCC-AR4). The performance index, \(I^2\), is calculated by averaging the normalized error variances across multiple climate variables. The results show significant improvements in model performance from CMIP-1 to CMIP-3, with the best models approaching the realism of atmospheric reanalyses. The superior performance of CMIP-3 models is attributed to more realistic parameterizations and finer resolutions, rather than just the use of more realistic forcing. The study also investigates the sensitivity of the performance index to the choice of variables and the robustness of the multimodel mean. The multimodel mean often outperforms individual models, suggesting that ensemble averaging can improve model predictions. However, the use of flux correction in earlier models is noted to have introduced systematic biases, which have been reduced in more recent models. In conclusion, while current models are not perfect, they are much more realistic than their predecessors, and the improvements in model performance and physical formulation suggest an increasing level of confidence in model-based climate predictions. The study highlights the importance of further research to develop more robust metrics for model evaluation and to address remaining limitations in model validation.The article by Thomas Reichler and Junsu Kim evaluates the performance of coupled climate models in simulating today's climate. Coupled climate models are sophisticated tools used to understand and predict climate change, but they are not perfect due to incomplete theoretical understanding and simplifying assumptions. The study aims to objectively quantify the agreement between model and observations using a single performance index derived from a broad group of variables. This approach is novel compared to previous model intercomparison studies, which often focused on specific processes or considered a narrow range of models. The study includes model output from three generations of climate models: CMIP-1, CMIP-2, and CMIP-3 (IPCC-AR4). The performance index, \(I^2\), is calculated by averaging the normalized error variances across multiple climate variables. The results show significant improvements in model performance from CMIP-1 to CMIP-3, with the best models approaching the realism of atmospheric reanalyses. The superior performance of CMIP-3 models is attributed to more realistic parameterizations and finer resolutions, rather than just the use of more realistic forcing. The study also investigates the sensitivity of the performance index to the choice of variables and the robustness of the multimodel mean. The multimodel mean often outperforms individual models, suggesting that ensemble averaging can improve model predictions. However, the use of flux correction in earlier models is noted to have introduced systematic biases, which have been reduced in more recent models. In conclusion, while current models are not perfect, they are much more realistic than their predecessors, and the improvements in model performance and physical formulation suggest an increasing level of confidence in model-based climate predictions. The study highlights the importance of further research to develop more robust metrics for model evaluation and to address remaining limitations in model validation.