Google Launches Gemini 4 Carbon with Advanced AI Capabilities

Google has officially launched Gemini 4 Carbon, its most advanced AI model to date, making it available to developers, enterprises, scientists, and consumers. The model, which builds on the capabilities of Gemini 4 Argon, is designed to tackle complex, multi-step problems in cybersecurity, software engineering, and scientific research, among other fields. According to Google, Carbon delivers "frontier performance" in long-horizon agentic workflows and reasoning tasks.

Key Features and Performance

Gemini 4 Carbon excels in tasks requiring deep reasoning and extended verification loops. It supports larger token limits than Argon, enabling thorough exploration and complex problem-solving. In independent benchmarks, Carbon has achieved top rankings, including #1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench, while maintaining over 2X lower cost per task compared to competing external models.

For cybersecurity, Carbon leads on benchmarks such as SEC-bench Pro, ExploitGym, and CWE-bench v1, demonstrating its ability to identify, exploit, and patch vulnerabilities. Additionally, it is highly effective in agentic workflows, such as coding, debugging, and scientific discovery, making it a versatile tool for developers and enterprises.

Applications in Multimodal and Agentic Workflows

Carbon's multimodal capabilities allow it to synthesize insights from dense documents, charts, and schematics. In collaboration with Artificial Analysis, Google developed the Multimodal Knowledge Agent (MKA) benchmark, where Carbon ranked #2 while maintaining a 6X lower average cost per task than the top-performing model.

Using Google Antigravity’s new /swarm feature, Carbon has demonstrated its ability to coordinate multi-agent workflows. For example, it reconstructed a 3D model of the Eiffel Tower's construction using historical photos and engineering plates, and built an interactive textile visualizer based on academic research. These projects highlight its capacity for innovative artifact generation and data analysis.

Pricing and Availability

Gemini 4 Carbon 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. It is also accessible to Google AI Ultra subscribers through Google Antigravity, the Gemini app, and AI Mode in Search.

With its advanced capabilities and competitive pricing, Gemini 4 Carbon is positioned to support a wide range of high-precision knowledge work, from legal and financial analysis to full-stack development and 3D design.