Google has launched Gemini 4 Carbon, its most advanced AI model yet, designed for developers, enterprises, scientists, and consumers. Carbon excels in complex tasks like cybersecurity, long-horizon agentic workflows, and scientific reasoning. It outperforms competitors on key benchmarks such as CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench, achieving over 2X lower cost per task compared to the next best model.
The model is particularly strong in high-precision knowledge work, artifact generation, and data analysis, ranking #2 on the new Multimodal Knowledge Agent (MKA) benchmark while being more than 6X lower in average cost per task. It also leads in agentic coding, full-stack development, and 3D/UI design, handling multi-agent collaboration and complex architectural redesigns.
In cybersecurity, Gemini 4 Carbon delivers cutting-edge capabilities, excelling in vulnerability detection, exploitation, and patching. It achieved 100% recall on scanning and 95% precision on triage in Google’s Project Fortify, and helped uncover critical vulnerabilities in open-source software. The model is available via the API in Google AI Studio and Gemini for business at the same introductory price as Gemini 4 Argon ($2 per million input tokens and $10 per million output tokens).
Google is also previewing /swarm in Google Antigravity, allowing developers to coordinate multiple agents for large, complex projects. Carbon is rolling out to all Google AI Ultra subscribers in various platforms, including the Gemini app and AI Mode in Search.