Google took a step that will intensify the competition in the field of artificial intelligence and presented the preview of its new generation artificial intelligence model, Gemini 2.0, to users. Sundar Pichai, the company’s CEO, stated in his blog post that Gemini 2.0 is their most capable model yet. Google Gemini 2.0 comes with an algorithm that natively supports image and audio output. Pichai said, “This model will make it possible to create new artificial intelligence agents that will allow us to get closer to our vision of a universal assistant.”
Google is following an interesting strategy in the launch of Gemini 2.0. Instead of the most advanced version, Gemini 2.0 Pro, it primarily offers the 2.0 Flash model, which is a more efficient and economical option. Starting today, Gemini users can access this version of the model via the web client. It is stated that use via mobile application will be activated soon.
Google aims to especially improve search services with Gemini 2.0. The new model will enable the AI Overviews feature in Search to answer more complex, multi-step math and coding questions. In addition, this feature, whose scope was expanded in October, will be available in more languages and countries.
Gemini 2.0 is not just limited to Search. The model is also being integrated into Google’s advanced artificial intelligence projects. These include Project Astra, a multi-modal AI agent introduced at the 2024 I/O event. Thanks to the new model, Astra can communicate in multiple languages and switch between languages instantly. It also offers features such as longer-term memory, low latency, and access to tools such as Google Lens and Maps.
Google Gemini 2.0 Flash’s performance surpasses previous models
Gemini 2.0 Flash goes beyond previous models in terms of performance. For example, it scored 63 points on the HiddenMath benchmark test, which measures its ability to solve competitive math problems, beating Gemini 1.5 Flash’s 47.2 points. Even more remarkable is that the experimental Gemini 2.0 version even surpasses the Gemini 1.5 Pro model in some areas. However, it lags behind the 1.5 Pro in areas such as long context understanding and automatic speech translation.
Due to this difference, Google will continue to use the Gemini 1.5 Pro model for a while. The company also announced a new tool called “Deep Research”. This tool can create comprehensive reports on complex topics using Gemini 1.5 Pro’s long context capabilities.