Google opens Gemini 4 Carbon to everyone: the strongest Gemini targets agent workflows and cybersecurity

#ai #cybersecurity #gemini #google

📅 November 10, 2026 • Google DeepMind

Per Google, Gemini 4 Carbon became publicly available on November 10, 2026 for developers, enterprises, scientists and consumers. Two weeks earlier, the model had been released to cybersecurity defenders through the Fairwind Program, where Google partnered with impacted organizations to remediate in record time, helping mitigate potential impact on tens of thousands of companies. Carbon is the second Gemini 4 frontier model, following Gemini 4 Argon, and is described by Google as its strongest model, with frontier performance in cybersecurity capabilities and in long-horizon agentic workflows across complex knowledge work, software engineering and scientific reasoning. In a multi-agent research setup, Google reports, Carbon autonomously discovered remote homologs across petabase-scale unannotated sequence archives.

The rollout covers consumer and professional surfaces at once. Google AI Ultra subscribers get Carbon in Google Antigravity, the Gemini app and AI Mode in Search. Developers can build on the Gemini API paid tier in AI Studio and Android Studio, while enterprises get access through Gemini Enterprise, including CodeMender. Per n8n, Carbon is available day-one in the n8n Google Gemini Chat Model node and in the standalone Google Gemini node. Scientists and researchers can apply for free access to both Argon and Carbon via the highest usage tier of Google Antigravity for 6 months.

Per Google via n8n, introductory pricing is 2 dollars per million input tokens and 10 dollars per million output tokens, rising to 4 dollars and 20 dollars after the introductory period ends on November 30, 2026. Published benchmarks are specific: 53.9 percent on AutomationBench for automation tasks, 88.2 percent on SEC-bench Pro for finding vulnerabilities, 54.8 percent on ExploitGym for checking exploitability and 71.0 percent on CWE-bench for patching. On the Gray Swan prompt injection benchmark Carbon showed a 0.0 percent attack success rate, down from 0.7 percent for Argon, though n8n notes no model is fully resistant at scale. Google also says internal teams have seen Carbon run for days on complex engineering problems, pushing through dead ends. Not every comparison is favorable: per The Verge, on the benchmarks Google cited at the first announcement Carbon sits behind Fable 5.2, and the outlet notes companies can cherry-pick benchmark results.

How it works in practice. For teams building automation on n8n, day-one node support means Gemini 4 Carbon drops into existing agent workflows without connector work, and the introductory token pricing holds until November 30, 2026. For security-oriented pipelines, the Gray Swan result of 0.0 percent is a data point for prompt injection risk reviews, with n8n's caveat that resistance at scale is not guaranteed. For vendor comparisons, The Verge's note about cherry-picked benchmarks is a reminder to ask which benchmark set stands behind a marketing claim.