Methods and apparatus for improving cache hits in network processors using flow-based packet assignment to compute engines. Packet processing operations are performed on a network processor having multiple compute engines via execution of instruction threads on those compute engines. Via execution of the threads, a flow-based packet processing assignment mechanism is implemented that causes at least a portion of the packet processing operations for packets associated with common flows to be performed on compute engines assigned to perform packet processing operations for those flows. This results in the same compute engines performing packet processing on packets assigned to common sets of flows, thus increasing the cache hits on data that is stored locally on the compute engines pertaining to the flows.