Computational models play central roles in the politics of the Anthropocene. Whether used to design seawalls or to project climate futures, we encounter them across a variety of sites of response to anthropogenic climate change and other environmental ills. In this paper, we reflect on the limits of contemporary computational modeling, particularly when it comes to representing coupled socioecological systems or human-nonhuman relations. Modeling practices naturalize specific assumptions about what counts as data, can be represented, and counts as legitimate knowledge regarding environments and their contents. They are also shaped by and reinforce contingent forms of computational logic and infrastructure that could, we suggest, be otherwise. As a result, their projections frequently grind against other ways of imagining environments and their futures, such as those developed through long-term inhabitation or embedded within different ecologies and infrastructures of knowledge production. They also materialize in infrastructures that can be part of the problem rather than the solution. Assembling insights from multiple ethnographic field sites, this paper maps a variety of practices that contest models and their limits. We ask both what models and their architectures assume and leave out (or cannot account for), the limits of their presumed ability to create social or political change, their materialization in infrastructures, and what strategies our interlocutors use to render visible what doesn’t fit their representational schemas. Our aim is to develop a framework for understanding the consequences of the limits of computational modeling from anthropological and global perspectives and the kinds of politics remodeling sets in train. Modeling environments and their futures not only precipitates political struggle between interest groups, we conclude. In the “pluriverse” that we inhabit, struggles over models and their limits are also clashes between heterogeneous ways of worlding, the results of which have important implications for whose “worlds” count, and how.