DALL·E 3 is a synthetic intelligence system that takes a textual content immediate as an enter and generates a brand new picture as an output. DALL·E 3 builds on DALL·E 2 by enhancing caption constancy and picture high quality. On this system card, we share the work finished to arrange DALL·E 3 for deployment, together with our work on exterior knowledgeable pink teaming, evaluations of key dangers, and mitigations to scale back the dangers posed by the mannequin and scale back undesirable behaviors.