ChatGPT was the first widely used AI chatbot and is still considered the top dog. But Google wants to knock the pioneer off its throne with a new version of its AI model Gemini. — Photo: Andrej Sokolow/dpa
SAN JOSE: Google has unveiled Gemini 3, a new AI model that will also power the company’s search engine and can generate interactive graphics to illustrate answers to scientific or mathematical questions.
Launching first in the search engine’s AI mode, Gemini 3 is better at processing and contextualising information across different media types such as video and text, Google said, announcing the update on Tuesday.
The company demonstrated the software using a camera feed to observe a pickleball game, analyse a player’s behaviour and offer tips for better results.
Google also said Gemini 3 could translate recipes from photos and then draft a cookbook. It also showed the software designing a programme for a multi-day trip to a city such as Rome, a demo commonly used by rivals.
Google says Gemini 3 is also better at coding, a frequent AI use case. To this end, the company introduced a new tool called Antigravity, which allows developers to deploy and supervise multiple largely autonomous AI programs, or agents, to write software.
Google recently said the Gemini app has around 650 million monthly users. By contrast, ChatGPT has more than 800 million users per week, according to OpenAI chief Sam Altman.
Google's advantage, however, is that its broader ecosystem of services is embedded in many people’s daily routines through tools like Search, Gmail and Maps.
For Search, Google says Gemini 3 does a better job than before of understanding the intent behind a user’s query, and of retrieving matching information from the web. – dpa
