- By Vikas Yadav
- Tue, 14 May 2024 11:05 PM (IST)
- Source:JND
Google I/O 2024 Updates: Google has announced 'Gemini 1.5 Flash' at its I/O event today. According to Demis Hassabis, Co-Founder and CEO of Google Deepmind, the model is a lightweight and cost-effective option next to other Gemini iterations. Gemini 1.5 Flash is trained efficiently for "narrower or high-frequency tasks" that are time-bound.
The company also updated Gemini 1.5 Pro at the tech show. Gemini 1.5 Pro is updated to include enhancements for use cases such as coding, translation, reasoning and more. These changes will be available from today onwards. Google's Gemini 1.5 Pro and 1.5 Flash will be available in over 200 countries and territories in preview starting today.
The general availability of the models will begin in June, Google said. These iterations of Gemini come with a one million token context window and will let users "interleave text, images, audio and video" as input prompts. Secondly, users can sign up for a waitlist in Google AI Studio or Vertex AI (for Cloud users) to gain access to the private preview of Gemini 1.5 Pro's two million context window.
Introducing Gemini 1.5 Flash ⚡
— Google (@Google) May 14, 2024
It’s a lighter-weight model, optimized for tasks where low latency and cost matter most. Starting today, developers can use it with up to 1 million tokens in Google AI Studio and Vertex AI. #GoogleIO pic.twitter.com/I1adecF9UT
For the unaware, context window size indicates the amount of data a model can process at a time. According to Hassabis, Gemini 1.5 Flash excels in tasks such as generating image and video captions, summarisation, chat applications and extracting data from documents, graphs, tables and more, according to Moneycontrol.
It has been trained via 'distillation', a process that transfers the basic skills and knowledge base from a bigger model into a smaller trim, which may also be more efficient. Meanwhile, Google also unveiled further additions (PaliGemma and Gemma 2) to the Geema Family.
PaliGemma (available today), a vision-language model, can generate image captions, answer questions based on visuals, and can perform other image labelling works. Moreover, the revamped Gemma 2 will launch in June. This comes amid other generative AI tools that the company announced today. You can read more on this here.