Google Gemini — status

Google Gemini is a family of multimodal large language models developed by Google DeepMind, representing a major advancement in artificial intelligence. It was created as a successor to Google’s previous models, LaMDA and PaLM 2, with the core goal of being inherently multimodal, meaning it was trained from the ground up to understand and operate across different types of information, including text, images, video, audio, and code. This makes it different from other models that have been retrofitted with multimodal capabilities after being initially trained on text alone. The Gemini family comes in different sizes, each optimized for specific tasks and devices. The most powerful version is Gemini Ultra, designed for highly complex, compute-intensive tasks like advanced reasoning, coding, and scientific research. It is the model that powers the premium Gemini Advanced service, which is a key part of Google's business model for this technology. The Gemini Pro model is a versatile, balanced version built for a wide range of tasks and is the model that powers the majority of the Gemini chatbot experience in many Google products. It offers a strong blend of performance and efficiency, making it suitable for a variety of general-purpose applications. The smallest and most efficient model is Gemini Nano, which is specifically designed to run directly on-device, such as on a smartphone, without a data network connection. This allows for tasks like on-device summarization, image analysis, and smart replies to be handled locally, improving speed and privacy. The business model for Gemini is multi-faceted. For individual users, it is offered through a tiered subscription model, with the free tier providing access to Gemini Pro and a premium tier, Gemini Advanced, which includes access to the more powerful Ultra model as part of the Google One AI Premium plan. For developers and enterprises, Gemini is available through APIs on Google AI Studio and Google Cloud Vertex AI, with a pay-as-you-go pricing structure based on the number of tokens consumed for both input and output. The competition in the large language model space is fierce, and Gemini is a direct competitor to OpenAI’s GPT series. While both models are highly capable, they have different strengths; Gemini's native multimodality gives it a powerful advantage for tasks that require seamless reasoning across various media types. It has been integrated into a wide array of Google's own products and services, including Google Search, Gmail, and Google Maps, to enhance their functionality and user experience.