Google’s Gemini 2.0 Family of Models Gets its First Member
Making waves across the technology sphere is almost a norm for the Internet giant, considering it recently introduced Willow, a quantum computing chip. What’s new is that Google’s Gemini 2.0 family of models is getting its first member - Gemini 2.0 Flash. It has been defined as a new AI model for the agentic era wherein agentic is a rough reference to having a better understanding of the world around users, thinking way ahead, and taking actions on their behalf.
Gemini 2.0 builds on the success of 1.0 and 1.5. Access to the experimental mode of Gemini 2.0 Flash is currently restricted to developers and trusted testers. The objective is to get productive feedback so that it can be integrated across Google products.
A hint is that the initial integration could be with Gemini followed by Search. There is a high possibility that the updated AI model will venture into the research segment, helping users get all the information they need in the most precise manner possible.
A stark comparison, for now, is being made with Gemini 1.5 Flash and Gemini 1.5 Pro.
One can briefly understand that Google’s Gemini 2.0 Flash is the company’s most popular model yet for developers. It comes with enhanced features, faster response time, and outperformance on key benchmarks.
Flash takes forward the multimodal support from input to output. Supporting multimodal output entails elements like image generation with text along with steerable TTS multilingual audio with TTS being an acronym for text-to-speech. Another advanced feature is calling tools like Google Search, code execution, and third-party user-defined functions natively.
Many more capabilities along with benchmarks have been listed by Google in its blog post. One would see a side-by-side technical comparison backed by marginal outperformance by Gemini 2.0 Flash Experimental in every category.
Gemini 2.0 is currently available for developers through Gemini API in Google AI Studio and Vertex AI. December is likely to be all about collecting feedback from participants so that Google can make improvements before the AI product goes live for general users. That is tentatively scheduled to happen in January 2025 and could be coupled with more model sizes.
For developers, Google has rolled out a new Multimodal Live API with real-time AV streaming input. Developers can also use multiple tools in their preferred combination. This will support working on Gemini 2.0 Flash Experimental.
Google’s Gemini 2.0 Flash Experimental is in its early stages. More developments would follow based on the feedback developers and trusted testers submit to the Internet giant.
Sundar Pichai talked about it in the same blog post, saying that Gemini 1.0 and 1.5 were successful in driving big advances. He added that Gemini 2.0 has been introduced with new advances in multimodality and native tool use.
Updates on Gemini 2.0 are coming after a decade-long investment by Google in its differentiated full-stack approach to AI innovation. Google’s Gemini 2.0 Flash Experimental has a lot of hopes riding on it, especially when it comes to being useful or productive to users.
Buy me a pizza if you loved this article.
Comentarios