Compressed natural gas vehicles are often considered a cleaner alternative to gasoline and diesel vehicles because they generally emit less particulate mass. However, their emissions of ultrafine particles and their potential biological effects remain insufficiently understood, especially for modern light-duty vehicles under realistic driving conditions. In this study, exhaust emissions from a Euro 6 compressed natural gas taxi were investigated on a chassis dynamometer using two driving cycles: a moderate real-driving cycle and a more dynamic cycle including cold-start operation. During the campaign, the vehicle exhibited two operating states: an initial rich-mixture condition associated with impaired aftertreatment performance, and a later stabilized condition. Gaseous pollutants, particle number, particle size distributions, soot mass, particle mass distribution, and deposited particle dose were measured. The nanoparticle-enriched particle fraction was chemically analysed for polycyclic aromatic hydrocarbons, nitrated and oxygenated derivatives, and water-soluble elements. In vitro toxicity was assessed using human lung epithelial cells exposed at the air–liquid interface to diluted gas phase and diluted whole exhaust. Rich-mixture operation strongly increased gaseous and particle emissions, while cold-start and dynamic driving also increased emissions under stabilized operation. The nanoparticle-enriched fraction contained low concentrations of organic compounds but substantially higher concentrations of water-soluble elements, dominated by zinc. Exposure to diluted exhaust reduced cell viability, increased membrane damage, and induced cytokine release. The gas phase alone produced measurable responses, while whole exhaust often produced stronger effects. However, differences between vehicle operating states and driving cycles were not consistent across all toxicological endpoints. These results show that the potential health relevance of compressed natural gas exhaust cannot be evaluated using particulate mass or regulated emissions alone. Even low-mass nanoparticle emissions, together with gas-phase compounds and soluble particle-associated species, may contribute to cytotoxic and inflammatory responses.