Economic decisions often require weighing multiple dimensions, or attributes. The orbitofrontal cortex (OFC) is thought to be important for computing the integrated value of an option from its attributes and comparing values to make a choice. Although OFC neurons are known to encode integrated values, evidence for value comparison has been limited. Here, we used a multi-attribute choice task for monkeys (Macaca mulatta) to investigate how OFC neurons integrate and compare multi-attribute options. By representing attributes with separate cues and using eye tracking to measure attention, we demonstrate that OFC neurons encode the value of attended attributes independent of other attributes in the same option. Encoding was negatively weighted by the value of the matching attribute in the other option, consistent with a comparison between like attributes. These results indicate that OFC computes comparisons between attributes rather than integrated values, and does so dynamically, shifting with the focus of attention.