American technology major Google launches Gemini 1.5 Pro with an expanded context window of two million tokens and made it available for developers in private preview.

“Gemini 1.5 Pro, with a one million token long context window, is now available with Gemini Advanced in 35 languages,” says Sundar Pichai, chief executive of Google and its parent company Alphabet.

Later this year, the business will introduce two million tokens to Gemini Advanced, enabling the upload and analysis of dense assets such as video and lengthy code, according to Sissie Hsiao, vice president, Gemini experiences.

Tokens are the building pieces that make up an AI model's "context window"; they are used to process information. Tokens can be single words, sentences, paragraphs, photos, videos, audio files, or computer code. A language model can analyze and absorb more information in each prompt if its context window is larger.

“Today, more than 1.5 million developers use Gemini models across our tools. You’re using it to debug code, get new insights, and build the next generation of AI applications,” adds Sundar Pichai.

Through the Gemini Multimodal Foundation Model app for iOS and Android devices as well as Gemini Advanced, which grants access to its models, users can engage directly with the model. To try it in three months, more than a million people have signed up thus far.

Processing power is needed for training language models. According to Pichai, the industry's need for machine learning computing has increased tenfold annually over the past six years, with a factor of one million growth. Google's fourth and fifth-generation tensor processing units were used to train Gemini.