Google has unveiled a public preview of Gemini 1.5 Pro, a powerful AI model capable of handling up to a million tokens, surpassing all previous AI models in terms of capacity. The release of this model was previously announced by Google.
Business users can access Gemini 1.5 Pro through Google's business service Vertex AI, as highlighted at the Cloud Next conference according to Google's announcement. Although the exact timeline for consumer availability is not specified.
This upgraded version boasts the ability to swiftly generate videos from 1-hour-long code files containing 30,000 lines of code or texts exceeding 700,000 words in just one processing session. Additionally, Google is currently developing a version capable of handling ten million tokens simultaneously, a far cry from the standard version's 128,000 token limit in Vertex AI.
It is anticipated that Gemini 1.5 Pro will match the performance level of the previous generation's most advanced model, Gemini 1.0 Ultra, as Google reported back in December. Another notable feature is the incorporation of a Mixture-of-Experts architecture, comprising multiple smaller models, each serving as an 'expert' in a specific domain. Based on the input prompt, the language model activates the most relevant expert(s) to craft a response, resulting in more efficient processing of prompts compared to utilizing the entire language model for each query.
English (US)