Google's Gemini 4 Carbon promises 'significant leap' in AI coding and cybersecurity (www.engadget.com)

Google has announced its latest AI model, Gemini 4 Carbon, which reportedly achieves significant advancements in coding and cybersecurity, surpassing OpenAI's Claude Opus 5.5 in key benchmarks. This new model builds upon the recently released Gemini 4 Argon and marks a rapid evolution in Google's efforts to enhance AI capabilities. Key highlights include a top score of 59 on the AA Index v5.0 benchmark, alongside strong performances in areas such as agentic coding, knowledge work, and scientific applications. Notably, Gemini 4 Carbon excels in cybersecurity, improving vulnerabilities identification and patching processes, all while aiming for enhanced model safety and reduced risk of exploitation. The introduction of Gemini 4 Carbon is particularly significant for the AI/ML community as it reinforces the competitive landscape of frontier models, showcasing Google's commitment to advancing secure coding practices. The model boasts a twofold increase in generating secure code compared to its predecessor, Argon. Initially, access to Gemini 4 Carbon will be restricted to select cybersecurity companies for pre-release testing, with broader availability planned through Google's Fairwind Program. This strategic rollout reflects Google's measured approach to deploying cutting-edge AI technologies while ensuring reliability and safety in critical applications.