Mountain watersheds are critical regions for water resources globally, and the timing and fate of the water they provide depends strongly on the temperature of the near surface atmosphere. We synthesize results showing that many regionally relevant climate models produce two-meter air temperatures (T2m) colder than what is observed (a “cold bias”), particularly in snow covered complex terrain during winter. We find common cold biases in 44 separate studies evaluating models across the world’s mountains, including single-model and multi-model ensembles. Using data from the SAIL field campaign, high-resolution WRF simulations, gridded observation based T2m datasets, and geographic analyses, we explore possible causes of these biases and examine the physical mechanisms, data limitations, and observational uncertainties that underlie our ability to model and measure T2m in complex mountain terrain. We determine that cold biases are likely a true bias, but also the sign of the biases can depend on the terrain-classification— in many cases, mountain valleys are too warm, whereas peaks are too cold. We conclude by emphasizing the value of coordinated model evaluation and field campaign development efforts in heavily instrumented mountain locations for addressing the root cause(s) of T2m biases and improving predictive understanding of mountain climates.