Google Expands Access to Gemini 4 Carbon Model

AI · First reported 10 Nov, 17:08 · Updated 17:31

Image: Wikimedia Commons · Francis Matahum · CC BY-SA 4.0

Google has expanded access to its Gemini 4 Carbon frontier artificial intelligence model beyond initial cybersecurity specialists to developers, enterprises, and Google AI Ultra subscribers. The model features enhanced capabilities in coding, workflow automation, and cybersecurity defense, alongside integration into platforms like Google AI Studio, Android Studio, and Antigravity.

Google began rolling out its Gemini 4 Carbon model on November 10 to developers, businesses, and Google AI Ultra subscribers across platforms such as the Gemini app, Search AI Mode, and Antigravity.
The release broadens availability two weeks after the model was initially offered strictly to a select group of cybersecurity defenders in late October.
Gemini 4 Carbon is designed for complex tasks including agentic coding, mathematics, science, and cross-application debugging, achieving first-place results on benchmarks like CursorBench 4.0 and AutomationBench.
On security benchmarks, Carbon recorded a 0% prompt-injection attack success rate on the Gray Swan IPI test, mitigating risks of cyberattacks and misuse.
Access for consumers requires a $100 monthly Google AI Ultra plan, while standard developer API pricing starts at $2 per million input tokens and $10 per million output tokens before doubling on December 1.