Google Expands Gemini 4 Carbon AI Access For Developers, Businesses With Advanced Coding, Cybersecurity And Research Capabilities

Google has opened up access to its most advanced AI model, Gemini 4 Carbon, to developers, businesses, scientists, and consumers. The company unveiled Carbon in a blog post, highlighting its superior capabilities in cybersecurity and long-running AI workflows for complex knowledge work, software engineering, and scientific research.

Carbon is Google's most powerful model, boasting frontier performance in areas such as cybersecurity and extended reasoning required for tasks like coding, debugging, and scientific research. The model can collaborate with other AI agents, enabling different systems to tackle various aspects of a larger project. Gemini Carbon is accessible via the Gemini API within Google AI Studio and Gemini for business, with a subscription price comparable to that of Gemini 4 Argon, priced at USD 2 per million input tokens and USD 10 per million output tokens.

Google's internal benchmarks show that Carbon outperforms other models, achieving top results on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench. The model also excels in financial research, matching the best model tested on Hebbia's financial research benchmark and topping Balyasny Asset Management's financial intelligence benchmark. Cybersecurity testing revealed that Google's Project Fortify, utilizing Carbon, achieved a perfect 100% recall in vulnerability scanning and 95% precision in triage.

The expanded access to Carbon follows a limited release to selected cybersecurity defenders through Google's Fairwind program. The model will be available through AI Ultra subscribers in Google Antigravity, the Gemini app, and AI Mode in Search.