Gemini 4 Carbon’s headline gains come with a narrow access window

Google introduced Gemini 4 Carbon on October 28, describing it as its strongest model and emphasizing cybersecurity and long-running engineering work. The practical change for developers is not broad access: Google says it is starting with a subset of cyber defenders, then plans to expand through its Fairwind program before making Carbon available through product surfaces. That puts the announcement ahead of general availability, so most teams cannot yet assess the model on their own workloads.

Carbon follows Gemini 4 Argon by about two weeks, according to Google. The company says it built Carbon on Argon and reports gains in coding, agentic tasks, and scientific work. It also lists an introductory API price of $2 per million input tokens and $10 per million output tokens, matching Argon’s introductory price. Those are announced prices, not a useful cost comparison for teams until they can access the model and learn how long the introductory period lasts.

Google reports Carbon scores of 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. It also says Carbon improved by 89 Elo points on WebDev Arena, reaching fourth place. These results come from Google’s announcement; the company says the cybersecurity benchmarks were evaluated without guardrails. Ars Technica’s comparison notes that Carbon’s margins over competing models in Google’s published tests are small and that the available results do not establish whether Carbon is a new model or a post-training update to Argon. The benchmark claims therefore need independent reproduction before they settle comparisons.

For security and coding teams, the reported improvements are relevant because they focus on finding and fixing vulnerabilities, rather than only general question answering. Google says Carbon more than doubled Argon’s success rate at generating secure code on an internal benchmark, and reports a strong patch-generation result from its Chrome team; those are company claims, not independent evaluations. The staged access means outside developers cannot yet check performance, latency, or reliability against their own codebases. Google has not given a date for wider access, so the announced capabilities and prices are not yet enough to plan a production migration.