This study explores the design of high-performance platinum intermetallic nanoparticle fuel cell catalysts using machine-learning-accelerated screenings. Carbon-supported PtCo alloys are promising for low-platinum oxygen reduction reaction electrocatalysis in proton-exchange-membrane fuel cells, but balancing particle size and ordering degree is challenging. By leveraging machine learning, the researchers quantified the impact of chemical ordering on thermodynamic stability and identified that introducing Cu/Ni into PtCo can enhance ordering and improve both specific activity and active surface area. Experimental preparation of small-sized, highly ordered Pt2CoCu and Pt2CoNi catalysts resulted in a large electrochemically active surface area of ~90 m² gPt⁻¹ and a high specific activity of ~3.5 mA cm⁻². These catalysts showed enhanced performance in practical H₂–air fuel cells, demonstrating a high mass activity of ~3 A mgPt⁻¹ and good durability under accelerated testing. The study highlights the effectiveness of machine learning in accelerating the discovery of optimal catalyst compositions, leading to improved fuel cell performance.This study explores the design of high-performance platinum intermetallic nanoparticle fuel cell catalysts using machine-learning-accelerated screenings. Carbon-supported PtCo alloys are promising for low-platinum oxygen reduction reaction electrocatalysis in proton-exchange-membrane fuel cells, but balancing particle size and ordering degree is challenging. By leveraging machine learning, the researchers quantified the impact of chemical ordering on thermodynamic stability and identified that introducing Cu/Ni into PtCo can enhance ordering and improve both specific activity and active surface area. Experimental preparation of small-sized, highly ordered Pt2CoCu and Pt2CoNi catalysts resulted in a large electrochemically active surface area of ~90 m² gPt⁻¹ and a high specific activity of ~3.5 mA cm⁻². These catalysts showed enhanced performance in practical H₂–air fuel cells, demonstrating a high mass activity of ~3 A mgPt⁻¹ and good durability under accelerated testing. The study highlights the effectiveness of machine learning in accelerating the discovery of optimal catalyst compositions, leading to improved fuel cell performance.