Google unveils Gemini 4 Carbon, first offering its top AI model to cyber defenders
Just two weeks after Gemini 4 Argon, Google has released Carbon, a model built to find and fix software flaws. Access is initially limited to selected security experts.
Archive image · Photo: Asoundd / Wikimedia Commons, CC BY-SA 4.0 (cropped and resized)
Google on Wednesday unveiled Gemini 4 Carbon, its newest advanced artificial intelligence model, about two weeks after launching Gemini 4 Argon. The company said Carbon is built for cybersecurity and complex, long-running coding work, and achieves top performance at finding, exploiting and patching software vulnerabilities. In internal tests, it produced secure code at more than twice the success rate of Argon, Google said.
The model is not yet widely available. Access goes first to a subset of cyber defenders through Google's Fairwind Program while the company continues testing safeguards. Google gave no date for a broader release. Introductory pricing matches Argon at $2 per million input tokens and $10 per million output tokens.
The launch intensifies Google's race with OpenAI and Anthropic, which also offer AI models for coding and security work. Constellation Research said the quick succession of releases shows Google speeding up its frontier model cadence. The benchmark results Google cited have not been independently verified.