CM3leon by Meta

CM3leon by Meta is a multimodal foundation model capable of both text-to-image and image-to-text generation. Designed for researchers and developers in AI, it features state-of-the-art performance on tasks such as image generation, captioning, visual question answering, and text-guided image editing. CM3leon stands out for its efficient transformer-based architecture, multitask instruction tuning across image and text, and strong zero-shot performance with less training compute compared to previous models.