Listen: Gemini
Google DeepMind's natively multimodal model family, announced December 2023, and the model line behind AI Overviews and AI Mode in Google Search.
Transcript
Gemini is Google DeepMind's flagship artificial intelligence model family, first announced in late 2023. Unlike older models that rely on separate add-ons to understand different types of media, Gemini is natively multimodal. It processes text, images, audio, and video all within a single system.
The family has evolved quickly to fit different needs. The first version came in three sizes, including a compact model called Nano that runs directly on mobile devices. Later versions introduced a massive memory span, capable of handling up to two million tokens of data at once, which allows users to prompt the model with entire libraries of information. Today, the lineup features Flash models built for speed and low cost, Pro models for high performance, and advanced versions with explicit reasoning capabilities.
You can access Gemini through Google's consumer app, their developer application programming interface, or enterprise platforms. But its biggest everyday impact is in Google Search, where it powers AI Overviews. Because Gemini decides what information to retrieve and cite, this single model family now determines which websites gain visibility in the modern search landscape.
