Spanish voiced obstruents are traditionally described as having a stop allophone [b, d, g] and a lenited allophone [β, ð, ɣ]. Despite this binary classification, acoustic data has shown that this variation is continuous or gradient depending on the preceding linguistic context. The goal of this paper is to investigate how the following linguistic context affects the degree of Spanish voiced obstruent lenition. Specifically, this paper reports an acoustic investigation of Spanish voiced obstruent lenition in onset cluster contexts. Nine native Spanish speakers were recorded reading Spanish-like nonce words that included a singleton voiced obstruent or an onset cluster consisting of a voiced obstruent plus [ɾ] or [l]. The relative intensity and the duration of these segments were measured and compared with linear mixed-effects regressions. In line with past work, the results show that the voiced obstruents are the most lenited in intervocalic contexts. However, Spanish voiced obstruents are significantly less lenited when followed by [ɾ] in a complex onset; when followed by [l] in a complex onset, the degree of lenition is much more variable. These results provide further support for the gradient lenition of Spanish voiced obstruents, rather than a dichotomous distribution of stops versus lenited variants.