Google launches Gemini 4 Carbon with a cybersecurity focus

Google has launched Gemini 4 Carbon, a new frontier artificial-intelligence model designed to tackle long-running coding and cybersecurity work, as competition intensifies with OpenAI and Anthropic. The model follows Gemini 4 Argon, which Google introduced about two weeks earlier.

Google said Carbon is initially available to a subset of cybersecurity defenders through its Fairwind Program. The company is gathering feedback and strengthening safeguards before extending access to developers, businesses and consumers. Carbon is not yet generally available.

Built for vulnerability discovery and repair

Google said Carbon handles complex, multi-step engineering tasks. Its published benchmarks put the model at 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation and 71.0% on CWE-bench for patching.

Google also said Carbon generates secure code at more than twice the success rate of Argon on internal benchmarks. Reuters reported the claim while noting that the comparisons come from Google’s own testing. Google said its teams have seen Carbon persist through difficult engineering tasks, including tracing and fixing subtle bugs across large codebases.

Pricing matches Argon

Gemini 4 Carbon is priced at $2 per million input tokens and $10 per million output tokens, matching Argon’s introductory rates, according to Google. Cached input tokens receive a 95% discount from the input-token price. Broader access is planned, but Google has not announced general availability across its consumer and enterprise products.

Sources: Reuters and Google.