Google recently made a series of Gemini announcements for developers at I/O 2024. One notable announcement was the introduction of a 2 million token contextual window for Gemini 1.5 Pro. This update allows the processing of 2 hours of video, 22 hours of audio, over 60,000 lines of code, and over 1.4 million words. Previously, Gemini Advanced with 1.5 Pro had only half this capacity. After a private preview, this feature is now available to all developers for use.
The expanded contextual window of Gemini 1.5 Pro has been beneficial for various organizations, including a fast food retailer, financial institution, insurance company, and a sports company analyzing a player’s swing. This advanced tool is useful for tasks such as error detection in extensive lines of code, information retrieval in research libraries, and analysis of audio and video content.
In addition to Gemini 1.5 Pro, Google also introduced Gemini 1.5 Flash, which offers a 1 million token context window, low latency, and competitive pricing. This version is ideal for retail chat agents, document processing, and research agents synthesizing entire repositories. Google highlighted the advantages of Gemini 1.5 Flash by comparing it to the current GPT-3.5 Turbo, emphasizing its larger token context window, faster processing speed, and lower entry price.
Google also unveiled Gemma 2, an open model available globally in parameter sizes of 9 billion and 27 billion. Furthermore, Image 3 was released in preview for Vertex AI early access customers, offering improvements such as faster generation, better understanding of instructions, photorealistic images of groups of people, and enhanced text representation within images.
An example provided by Google shows a photorealistic image of a woman’s hand reaching out to touch a dandelion seed in a field of dandelions, accompanied by the phrase “Sometimes letting go is the bravest act” written above the hand in delicate handwriting. These advancements in Gemini and Google’s other tools aim to provide developers with more efficient and effective capabilities for a variety of applications.
Article Source
https://9to5google.com/2024/06/27/gemini-1-5-pro-2-million/