Open-Source AI & Open Models Reading List

An AI model's openness involves several separate questions: whether people can download it, what its license permits, whether training data is accessible and what running it costs. Nathan Lambert's open-model reading list makes those distinctions a starting point for understanding the field, before moving into competition and training methods.

One foundational paper in the list proposes a spectrum of release methods, running from fully closed through hosted and downloadable access to fully open releases. That framework helps explain why two products described as open can give researchers or developers very different degrees of control. The list is useful as a route through those questions, rather than as a single ranking of models.