Google Announces Gemini 4 Carbon Frontier AI Model
Google has officially [announced Gemini 4 Carbon](https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon/), a new frontier AI model that builds upon the capabilities of the previously released [Gemini 4 Argon](https://arstechnica.com/google/2026/09/google-announces-gemini-4-argon-ai-model-but-yet-cant-use-it-yet/).
Radwan AbdulhadiEditor and technology writer
Introduction to Gemini 4 Carbon
Just two weeks after rolling out Gemini 4 Argon, Google has officially announced Gemini 4 Carbon as its latest frontier artificial intelligence model. While not the Flash model equivalent that some users have anticipated, Carbon serves as an incremental improvement over Argon. Similar to its predecessor, Google is utilizing a staged release approach for the new model, meaning it may take some time before everyday users can interact with it directly.
Google states that its latest AI model reaches a new threshold in coding, agentic tasks, and scientific analysis. The company positions Carbon as a totally new model, though it could potentially function as a post-training upgrade of Argon. Both models share substantial similarities, though Carbon outperforms Argon by slim margins across testing parameters provided by Google.
Cybersecurity and Performance Gains
Gemini 4 Carbon reportedly delivers major advancements in cybersecurity applications. According to Google, the model displays a significant leap forward in vulnerability detection, validation, and remediation. Furthermore, Carbon is supposedly twice as likely to generate code free of vulnerabilities compared to Argon. The Chrome team also claims that Carbon performs ten times better at generating high-quality patches than non-Google models.
In addition to cybersecurity improvements, Google notes that Carbon increased its score by 89 points on the WebDev Arena benchmark, landing it at the number four position on the leaderboard. Performance evaluations show Carbon slightly ahead of competitors such as GPT-6 Astra, Claude Opus 5.5, and Claude Fable 5.1 in nearly every tested category, although industry benchmarks often show frontier models scoring within the same general neighborhood.
Safeguards and Evaluation Results
Alongside its improved capabilities, Google integrated more rigorous safety safeguards into Carbon. During evaluations conducted by Grey Swan, Carbon successfully resisted every attempted prompt-injection attack. This 0 percent vulnerability score outperforms Argon at 0.7 percent, Opus 5.5 at 1 percent, and GPT-6 Astra at 8.5 percent.
Release Plans and Pricing Structure
The release schedule for Gemini 4 Carbon mirrors the rollout strategy utilized for Gemini 4 Argon. Because of the model's strong performance in agentic coding and cybersecurity, Google is employing a deliberative approach that complies with the voluntary testing framework established by the United States government. Once testing concludes, Google plans to distribute the model through its Fairwinds cyber defense program.
Google has not committed to a definitive timeline for a wide public release, though if it follows the pattern established by the Gemini 4 Argon release, the model could potentially become available in a couple of weeks. When it does launch, Carbon will likely arrive first for AI Ultra subscribers and software developers leveraging the Gemini API.
Similar to Argon, Google intends to offer a temporarily discounted API rate of $2 per million input tokens and $10 per million output tokens for Carbon. These promotional prices are scheduled to double once the introductory period expires. While Argon features fixed discounted rates through the end of November, Google has not yet specified an expiration date for Carbon's introductory pricing structure.
Sources
- Ars TechnicaGoogle announces Gemini 4 Carbon, its latest frontier AI model that you can't use
Continue chronologically
Related entity coverage
Related stories
Google Introduces Gemini 4 Carbon for Cyber Defense
Google has announced Gemini 4 Carbon, a new frontier AI model designed to find, exploit, and patch software vulnerabilities with advanced capabilities.
Google DeepMind Announces Gemini 4 Carbon
Google DeepMind introduces Gemini 4 Carbon, its strongest frontier AI model featuring advanced cybersecurity capabilities and long-horizon engineering performance.
Google Launches Gemini 4 Carbon Model With Cybersecurity Focus
Google has announced Gemini 4 Carbon, its latest frontier model featuring significant gains in coding, agentic tasks, and cybersecurity defense.
Google Announces Gemini 4 Argon AI Model
Google has announced its new Gemini 4 Argon AI model, claiming industry-leading performance in coding, knowledge work, and cyberdefense.
