Google has announced Gemini 4 Carbon, its newest frontier AI model, just two weeks after releasing Gemini 4 Argon. While not a revolutionary leap, Carbon shows significant improvements in coding, agentic tasks, and scientific analysis. The model reportedly excels in cybersecurity, detecting vulnerabilities more effectively and generating vulnerability-free code twice as often as Argon.
Chrome team claims Carbon is 10 times better at generating high-quality patches, boosting its Web Dev Arena score by 89 points to rank fourth overall. Despite its capabilities, Google hasn't made it publicly available, maintaining a staged release approach.
Source: Ars Technica · Summarized by HeadlinesBriefing