Will work for you continuously for days: Google's most powerful model revealed

Google has unveiled its most powerful AI model, Gemini 4 Carbon, ranked first in the world on the Artificial Analysis index. The model can operate continuously for days and solve complex engineering tasks, while collaborating with the Israeli company Wiz in the field of cybersecurity.

Google has surprisingly unveiled its new AI model, Gemini 4 Carbon, which CEO Sundar Pichai defined as the most powerful the company has ever built. The model ranks first in the world on the Artificial Analysis v5.0 AI index with 59 points, and leads in cybersecurity tests with 88.2% in identifying security vulnerabilities and 71% in bug fixing. The main innovation is its ability to operate as an autonomous agent continuously for entire days, analyzing massive code systems and identifying root causes. The Google Chrome team reported ten times higher efficiency in producing security patches compared to competitors. Collaboration with the Israeli cloud security company Wiz confirmed that the model leads by a significant margin in exposing real-world vulnerabilities. In safety tests, the model recorded 0.0% success for attackers in attempts to bypass instructions. Initial prices stand at $2 per million tokens for input and $10 for output, and after launch will rise to $4 and $20 respectively. In the first phase, the model is opened for use by a limited group of cybersecurity personnel.