On October 28, 2026, Google announced **Gemini 4 Carbon**, which it describes as its most powerful AI model, with state-of-the-art cybersecurity capabilities and stronger performance on long, complex tasks. It arrives just two weeks after Gemini 4 Argon, its previous model. ## What's new - **First place on the Artificial Analysis index.** According to Google, Carbon scores 59 on version 5.0 of the Artificial Analysis Intelligence Index, putting it in the top spot. - **Better coding and agents.** Google reports significant gains in coding tasks and in agents that work autonomously for longer. - **A focus on cybersecurity.** It leads specialized tests: SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploiting them in controlled tests, and CWE-bench (71.0%) for patching them. In internal tests, it also generates more secure code. ## Who can use it For now, not the general public. Google chose a gradual rollout: starting this week, Carbon is available to a group of cybersecurity defenders, and access will then expand to other trusted partners through its **Fairwind** program, which prioritizes governments, critical infrastructure operators and core technology platforms. Google is also taking part in the U.S. government's voluntary process for testing models before release. The company says it will bring Carbon to developers, businesses and consumers "as soon as possible" after refining its safeguards, but gave no date. ## How much it will cost Once available, Carbon will have an introductory price of $2 per million input tokens and $10 per million output tokens, with cached input tokens 95% off. After that period, pricing will be $4 and $20 per million tokens, respectively. ## Why it matters AI models no longer just write code: they also find and fix security flaws. Google giving defenders access first shows how big tech companies are trying to make these capabilities help protect software before they are used to attack it. For any business with a website or app, it's a sign that AI-assisted security reviews will become increasingly common. **Sources:** - Google, "Gemini 4 Carbon: our next frontier model" (October 28, 2026): https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon/ - Google DeepMind, "Fairwind Program": https://deepmind.google/fairwind-program/ - 9to5Google, "Google announces Gemini 4 Carbon with 'significant gains across coding and agentic tasks'": https://9to5google.com/2026/10/28/gemini-4-carbon-announcement/