Purpose: As toddlers, non-autistic children learn to prioritize the shape of objects when extending a new countable noun, but not a new adjective.  Since autistic children are typically delayed in language learning they do not learn an appropriate “shape bias” until they are older. Yet prior work has claimed that the shape bias poses a specific challenge, lasting until and unless autistic children’s vocabularies reach that of non-autistic 4-5-year-olds.  Methods: The current study tests a large cohort of autistic preschoolers (N = 102) and a control group of non-autistic 2-year-olds (N = 31) on novel noun and adjective extensions. Results: Included autistic children who understood and completed the task (n = 67) displayed equivalent expressive vocabulary and equally strong performance as non-autistic 2-year-olds. Autistic children excluded for failure to understand or engage with the tasks (n = 35), were found to comprise a distinct subgroup, with only ¼ the vocabulary and more autism characteristics than included autistic children. Six autistic children with somewhat lower shape bias performance showed significant improvement after an eight-week intervention. Conclusion: The shape bias does not pose a specific challenge to autistic children if and when their vocabularies reach that of non-autistic 2-year-olds. Thus, autism-related generalization challenges do not result in specific difficulty extending word meanings on the basis of single feature: shape.