Google announces Gemini 4 Carbon, plans limited cyber rollout

Google announced Gemini 4 Carbon on October 28, describing it as its strongest model and reporting top results on a widely used independent AI leaderboard. The company plans to give a small group of cybersecurity defenders access first, with broader availability to follow after further testing and feedback. Carbon scored 59 on version 5.0 of the Artificial Analysis Intelligence Index, one point ahead of Claude Opus 5.5 Max, which scored 58. The index combines results from multiple evaluations. Artificial Analysis also reported that Carbon led in four of the index’s nine individual evaluations. The scores describe performance on selected tests, not necessarily how the model will perform across all real-world tasks. The launch follows Google’s release of Gemini 4 Argon, which reached the Gemini app two weeks earlier. Google says Carbon improves on Argon in coding, agentic work, scientific tasks, full-stack web development and multimodal understanding. On WebDev Arena, a user-voted evaluation of AI-generated websites, Carbon placed fourth, with a reported score of 1,768. Google is emphasizing cybersecurity applications, including finding vulnerabilities and helping develop fixes....