During its Google I/O 2024 developer conference, Google LLC unveiled a suite of new artificial intelligence models and tools aimed at enhancing resources for developers engaged in AI and machine learning application development.
Highlighting the event, Google introduced Gemini 1.5 Flash, its latest and quickest model to date. Built on the Gemini Pro backend, Flash is tailored for accelerating high-frequency tasks and boasts capabilities similar to a superhero, excelling in speed.
Like its predecessor, Gemini Pro, the Flash model supports multimodal interactions, capable of processing and generating text, images, and audio. It also features a vast 1 million-token context window. In addition, the Gemini 1.5 Pro model is set to offer a private preview of a 2 million-token context window, doubling its current capacity, although developers will need to join a waitlist to gain access.
Both models are now accessible in over 200 countries and territories, including the U.K. and Switzerland.
Google also announced an expansion to its lightweight open model family, Gemma, inspired by Gemini. The new addition, named PaliGemma, is a large language model equipped with vision capabilities, suitable for tasks like image labeling and image-based Q&A. It can be utilized on various devices.
Furthermore, an upgrade to the Gemma series is underway with Gemma 27B, a more robust model that is currently in preview as part of Gemma 2, scheduled for release in June. Designed to operate on Nvidia Corp. GPUs or a single Google tensor processing unit, Gemma 27B promises performance that rivals models double its size.