Google DeepMind unveils gemini 4 carbon, a new frontier model with advanced capabilities

Google DeepMind has made Gemini 4 Carbon available to developers, enterprises, scientists, and consumers, positioning it as the company's most powerful model, particularly for complex tasks in cybersecurity, software engineering, and scientific research. While Gemini 4 Argon remains suited for general knowledge work, Carbon is designed for intensive tasks like deep debugging and open-ended scientific discovery, utilizing more tokens for thorough reasoning. Carbon demonstrates top-tier performance across several independent benchmarks, ranking first on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench, while maintaining a cost advantage of over two times lower than the next best external model. Furthermore, its specialized cybersecurity iteration leads on key benchmarks across finding, exploiting, and patching vulnerabilities when accessed through the Fairwind Program. The model is accessible via the API in Google AI Studio and Gemini for business at the same introductory price as Argon ($2 per million input tokens and $10 per million output tokens). Advanced features, such as the /swarm capability in Google Antigravity, allow Carbon agents to execute complex projects, including building interactive 3D models from historical documents and reconstructing historical instruments from X-rays. Additionally, Carbon is showing strong performance in translational science, with Axiom Bio adopting it for drug toxicity research.