Google has unveiled Gemini, its most capable and flexible AI model yet, designed to rival OpenAI’s GPT-4. The model is multimodal, meaning it can process and understand different types of information like text, images, audio, video, and code.

Gemini comes in three sizes: Ultra, Pro, and Nano. Gemini Ultra is the largest and most powerful model, intended for highly complex tasks. Google claims it outperforms current state-of-the-art models in many benchmarks. It is the first model to exceed human expert performance on the Massive Multitask Language Understanding (MMLU) benchmark, which uses a combination of 57 subjects such as mathematics, physics, history, law, medicine and ethics for testing both world knowledge and problem-solving abilities.

Gemini Pro is designed for a wide range of tasks and will be available to developers through the Google AI Studio and Google Cloud Vertex AI. It is being integrated into Google products like Bard immediately.

Gemini Nano is designed for on-device tasks on mobile devices. Pixel 8 Pro is the first smartphone engineered to run Gemini Nano directly on the device.

The company emphasizes Gemini’s ability to reason, understand, and combine different types of information. Google showcased Gemini’s capabilities with examples such as understanding complex math and physics problems from videos, answering questions about images, and generating code in multiple programming languages.

Google plans to continue improving Gemini and integrating it into more products and services in the future. The release of Gemini marks a significant step in the AI landscape, potentially impacting various industries and applications.