Google has announced Gemini 4 Carbon, which the company describes as its most capable AI model for long-horizon tasks, in a post on the Google blog on October 28. Access starts with a subset of cyber defenders this week. Developers, businesses, and consumers cannot use the model yet.
Gemini 4 Carbon arrives two weeks after Gemini 4 Argon. Google says that in internal testing, Carbon worked on engineering tasks for days, pushing through roadblocks to diagnose, root-cause, and patch subtle bugs across large codebases.
Gemini 4 Carbon Benchmarks and API Pricing
Google reports that Carbon ranks first on the Artificial Analysis Intelligence Index v5.0 with a score of 59. On WebDev Arena, the model gained 89 Elo points to reach 1768, which places it fourth.
Cybersecurity is the focus of the launch. Google lists these scores:
- SEC-bench Pro, which measures autonomous discovery of real-world vulnerabilities: 88.2%
- ExploitGym, which tests whether discovered flaws can be exploited: 54.8%
- CWE-bench, which measures turning validated flaws into working code patches: 71.0%
The company claims Carbon has more than twice the success rate of Argon at generating secure code in internal benchmarks. The Chrome security team also reported that Carbon produced high-quality patches at more than ten times the rate of another lab’s large frontier model. Google notes that the external cybersecurity benchmarks were evaluated without guardrails.
Introductory API pricing matches Gemini 4 Argon at $2 per million input tokens and $10 per million output tokens, with cached input tokens priced 95% below the standard input rate. After the introductory period ends, prices rise to $4 per million input tokens and $20 per million output tokens.
Gemini 4 Carbon Availability Through the Fairwind Program
The first wave is limited to defenders. Google says it will make Carbon available to a subset of cyber defenders starting this week, then phase in more trusted partners through its Fairwind Program.
The company is still “gathering feedback and iterating on guardrails” and plans to open access to developers, businesses, and consumers “as soon as possible.”
Carbon is built on the same safety stack as Argon, according to Google. In the company’s own testing, the attack success rate on the Gray Swan indirect prompt injection benchmark fell to 0.0%, compared with 0.7% for Argon, and high-severity misalignment alerts in live internal coding environments dropped by almost 80%. Google is also taking part in the US government’s voluntary pre-release model testing process.
Developers can plan budgets around the published introductory and standard prices now. Google has not given a date for public API access, and all benchmark figures come from the company’s own announcement without independent verification.
Thank you for being a Ghacks reader. The post Google Launches Gemini 4 Carbon for Cyber Defenders First, With API Pricing From $2 per Million Tokens appeared first on gHacks.
