Lower token consumption and reduced memory pressure can decrease context compression and hitting limits, thereby enabling more work to be completed on the same hardware.