Long‐term field experiments are useful for determining cropping system productivity, stability, and resource use efficiency. With 12 yr (2004–2015) of data from five cropping systems on a long‐term experiment (> 30 yr) under semiarid conditions in Saskatchewan, Canada, a systems‐approach was used to compare grain and protein yield, stability, nitrogen (N) dynamics, N fertilizer (FUE G,P ), and available N use efficiency (NUE G,P ) for grain and protein. Annualized grain and protein yields for wheat ( Triticum aestivum L.)‐canola ( Brassica napus L.)‐wheat‐field pea ( Pisum sativum L.; W‐C‐W‐P) were 2244 and 372 kg ha −1 , respectively, 14 to 38% and 33 to 66% higher, respectively, than continuous wheat (ContW), summer fallow‐wheat‐wheat‐wheat (F‐W‐W‐W), F‐W‐W, and lentil ( Lens culinaris Medik) green manure‐wheat‐wheat (GM‐W‐W). Fallow systems were the most stable, but less productive and well‐adapted to low‐yielding conditions, while GM‐W‐W was the least stable and poorly adapted. The ContW had below‐average stability and was better suited to high‐yielding conditions for grain. The W‐C‐W‐P consistently produced above‐average yields, and was best suited for high‐yielding conditions for grain and protein. The ContW and W‐C‐W‐P had the highest NUE G (26.4 g kg −1 ) and NUE P (4.1 g kg −1 ), respectively, with GM‐W‐W having the lowest (18.1 and 2.7 g kg −1 ); FUE was the reverse of NUE. This long‐term study showed that diversified cropping systems that include pulses can more consistently produce higher grain and protein yields, regardless of growing conditions, than most other systems with lower N fertilizer inputs, thereby potentially reducing the negative environmental consequences associated with N fertilizer application.