Against the backdrop of addressing climate change and promoting green low-carbon development, the carbon emission performance of the Yangtze River Basin is of great significance for China to achieve its dual-carbon goals. Taking 113 prefecture-level cities in the Yangtze River Basin from 2000 to 2022 as research objects, this study comprehensively employed the super-efficiency SBM model, spatial autocorrelation analysis, spatial Durbin model, and traditional/spatial Markov chains. From the dual scales of the entire basin and the sub-basins of the upper, middle, and lower reaches, it explored the spatial-temporal evolution characteristics, multi-scale influencing mechanisms, and long-term evolution trends of carbon emission performance. The results showed that: ① The total carbon emissions of the Yangtze River Basin exhibited a three-stage characteristic of “rapid growth—slowing growth—platform fluctuation”, with 2020 as the growth inflection point, and the spatial distribution followed the pattern of downstream > midstream > upstream. ② The overall carbon emission performance was relatively low, showing an “N”-shaped fluctuating evolution trend (rise—decline—rise) in the time series, and presented the spatial characteristics of low-value expansion, medium-high value contraction, and core-periphery differentiation, with significant positive spatial autocorrelation. ③ At the whole basin level, the number of green patents was the core influencing factor, while carbon emission intensity and high-carbon industrial structure inhibited the improvement of performance. The influencing mechanisms of the upper, middle, and lower reaches showed significant heterogeneity: carbon emission intensity was the key negative driving factor in the upper reaches, with low transformation efficiency of green patents; the inhibitory effect of carbon emission intensity was the strongest in the middle reaches, while insufficient implementation of environmental regulation and employment dependence on the energy industry exacerbated carbon lock-in; the number of green patents had a significant positive effect on performance improvement in the lower reaches. ④ Predictions based on the spatial Markov chain indicated that performance evolution presented characteristics of club convergence and neighborhood dependence. Both high and low performance had high self-maintenance probabilities; the upper and lower reaches faced the problem of high and low performance lock-in, while the middle reaches showed significant low-performance lock-in. This study provides a scientific basis for achieving the dual-carbon goals and advancing cross-regional collaborative governance in the Yangtze River Basin, so as to promote the high-quality development of the basin.