Software Google attacks ChatGPT with an improved AI model
In the race to develop artificial intelligence, Google is forging ahead with new software that is used in the company's search engine, among other things.
19.11.2025, 08:33
SDA
Among other things, the new Gemini 3 AI model can also create interactive graphics to illustrate answers to scientific or mathematical questions.
Among other things, the new Gemini 3 model is said to be better at processing information from different types of media such as video and text together and putting them into context. Google demonstrated this with an example in which the AI software observed a pickleball game via the camera and was able to analyze a player's behaviour and give advice for better results.
Gemini 3 could also translate recipes based on photos, for example, and then use them to create a cookbook. Google also demonstrated that the software could design a program for a multi-day trip to a city like Rome. This is a common example that competitors also use.
650 million users per month
Gemini 3 should also be better at programming software - a common field of application for AI. To this end, the company presented a new tool called Antigravity. This enables software developers to use and monitor several, largely autonomous AI programs - known as agents - to write software.
According to the latest figures from Google, the Gemini app has around 650 million users per month. According to OpenAI CEO Sam Altman, ChatGPT has more than 800 million users per week. However, Google is integrated into the daily routine of many people with its search engine and other services such as GMail. When it comes to search, Gemini 3 should be better than before at understanding the intentions behind a user's query - and searching the web for relevant information.
