In visual word processing modeling, few models have successfully accounted for a large variety of tasks, and large corpora of behavioral observations. We consider a dataset from a megas-tudy, in which participants performed three tasks (lexical decision , word naming, and word recognition in a progressive demasking situation), on the same, large set of stimuli. We define the BRAID-Phon model, an extension of a previous probabilistic model, the BRAID model, whose originality is its visuo-attentional component, in which a visuo-attentional distribution spatially deploys sensory processing capabilities. BRAID-Phon includes phonological representations of words, allowing simulating the naming task. We simulated the three tasks on the dataset we considered, and analyzed predicted reaction times in terms of word frequency and word length effects. Simulation results show that BRAID-Phon successfully captures the direction and order of magnitude of the observed effects, in all three tasks.