Google DeepMind Releases Gemini 4 Carbon

Google has released Gemini 4 Carbon, a new model built for tasks that require deep reasoning, multi-step coding, debugging, and open-ended scientific research.

The model is available through the API in Google AI Studio and Gemini for business for paying developers and enterprises. Introductory pricing is set at $2 per million input tokens and $10 per million output tokens. It is also rolling out to Google AI Ultra subscribers in Google Antigravity, the Gemini app, and AI Mode in Search.

According to Google, Gemini 4 Carbon ranks #1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench at more than 2X lower cost per task compared to the next external model.

For developers working on complex projects, Google is previewing the /swarm feature in Google Antigravity. This allows a single prompt to turn into a coordinated team of agents assigned to architectural design, coordination, and auditing roles.

Google states that when accessed through the Fairwind Program, the model leads on benchmark evaluations across finding vulnerabilities on SEC-bench Pro, exploiting them on ExploitGym, and patching them on CWE-bench v1.