We aimed to explore whether there were cold and heat temperature adverse effects on years of life lost (YLL) for non-accidental mortality in Yuxi, a southwest plateau region of China. From data for 89,467 non-accidental deaths over an 8-year study period, we used a general linear regression model combined with a distributed lag non-linear model to assess the burden of disease non-accidental mortality due to ambient temperature with the YLL indicator. We estimated the mean YLL change per 1 degrees C decrease from the 25th to 1st percentile mean temperature as the cold effect and per 1 degrees C increase from the 75th to 99th percentile as the heat effect. The 95% empirical confidence intervals (eCIs) were calculated by using a bootstrap simulation method. The exposure-response curve between average temperature and YLL was U-shaped. The cold effect peaked at the first day after exposure and disappeared at 2 weeks, and the heat effect only lasted for the first 3 days. A per 1 degrees C decrease from the 25th to 1st mean temperature percentile was associated with an increase of 15.6 (95% eCI: 2.4, 22.9) in YLL for non-accidental diseases, and the cumulative effects due to cold were stronger in contrast to that attributed by heat. Cold temperature had a significant impact on YLL among the subgroups, with higher YLL in cardiovascular disease, stroke, males, Han nationality, married, and those engaged in agriculture than their corresponding categories. An increasing death burden of non-accidental in Yuxi of China due to cold temperature was demonstrated, and the association was also modified by specific disease causes and individual features.