A large multimodal workhorse that combines strong language reasoning with genuine image understanding. It can describe scenes, interpret diagrams, and answer questions grounded in visual content — not just text. As an open-weight model, it trades some of the polish of closed systems for full transparency and local deployment flexibility.