Google has introduced Gemini 4 Carbon, a new frontier AI model it says is designed for complex coding work and cybersecurity defense. Announced on October 28, the model builds on Gemini 4 Argon and is aimed at longer engineering tasks, including finding software vulnerabilities and helping developers fix them.
Google reports that Carbon scored 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for testing whether flaws can be exploited, and 71.0% on CWE-bench for patching weaknesses. The company also says the model ranked first on the Artificial Analysis Intelligence Index version 5.0 with a score of 59. These results come from specific benchmark evaluations and do not guarantee the same performance in real-world environments.
Carbon is initially being made available to a subset of cybersecurity defenders through Google’s Fairwind Program. Google says access will gradually expand to additional trusted partners before the model becomes available to developers, businesses and consumers through its products. The company has not announced a date for the wider release.
Google says Carbon achieved a zero-percent attack success rate on Gray Swan’s Indirect Prompt Injection benchmark, compared with 0.7% for Gemini 4 Argon. The company also reports more than a twofold improvement in secure-code generation over Argon and an almost 80% reduction in high-severity misalignment alerts in its internal live-coding tests. These figures are based on Google’s own evaluations.
Introductory pricing matches Gemini 4 Argon at $2 per million input tokens and $10 per million output tokens. Cached input tokens are priced 95% below the standard input rate. Google says the price will rise to $4 per million input tokens and $20 per million output tokens after the introductory period.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
