Google has unveiled Gemini 1.0, a next-generation multimodal model that revolutionizes the standard approach to such models. In contrast to previous methods involving separate modality components stitched together, Gemini is natively multimodal, pre-trained on various modalities and fine-tuned with additional multimodal data.