Google has announced Gemini 1.5, the next version of its main language model. Gemini Pro 1.5 can handle a maximum of 1 million token inputs, four times as much as GPT-4 from competitor OpenAI. This allows users to issue more complex commands.
Gemini 1.5 Pro is According to Google Capable of instantly processing 1-hour videos, code files of 30,000 lines of code, or scripts of 700,000+ words in one go. Google is also working on a version that can process ten million icons at once. This is possible via Vertex AI: the standard version of Gemini 1.5 Pro has a limit of 128,000 tokens.
The Gemini 1.5 Pro should also roughly reach the level of the Gemini 1.0 Ultra, the more comprehensive model of the previous generation. Google announced this in December. Another novelty is the use Structure of a mix of experts. This consists of several smaller models, each an 'expert' in their own field. Depending on the prompt, the language model activates one or more of these experts to answer the prompt. As a result, for many prompts, prompt processing is more economical than if the entire language model had to be used for each prompt.
Gemini 1.5 Pro will be available to some business customers starting Thursday. A version for all business and consumer customers will be released later, but Google has not specified when. Gemini for phones will continue to run on Gemini 1.0 for the time being, including the paid version called Gemini Advanced.
“Lifelong entrepreneur. Total writer. Internet ninja. Analyst. Friendly music enthusiast.”
More Stories
Epic will build its own Epic Games Store: here are the details
An eye-catcher at PAD Paris 2024
Google Play Store now lets you download two apps at once