Google launches Gemini 4 Carbon with cyber, coding claims

Google made Gemini 4 Carbon generally available on November 10, pitching it as frontier-class in cybersecurity and complex knowledge work at introductory pricing of $2 per million input tokens and $10 per million output tokens. Those rates rise to $4 and $20 after November 30, 2026.

In the launch post, Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect, writes: "Today, we're making Gemini 4 Carbon available to developers, enterprises, scientists, and consumers." The company says Carbon "ranks #1 on CursorBench 4.0, Perplexity's WANDR, and Zapier's AutomationBench at more than 2X lower cost per task compared to the next best external model."

The cybersecurity framing is where the announcement pushes hardest. Google says Carbon "leads on key benchmarks across finding (SEC-bench Pro), exploiting (ExploitGym), and patching (CWE-bench v1) vulnerabilities," and reports that in internal testing it discovered 0-day vulnerabilities in major SaaS providers within two hours. On the defense side, Carbon "blocked 99% of attack trajectories from our adaptive red-teaming agent" and "achieved a 0% prompt injection attack success rate on the Gray Swan IPI benchmark," plus a 92% pass rate on a 375-task agentic honesty benchmark. The post does not name which SaaS providers were affected.

Alongside the launch, Google opened a Fairwind Program giving cybersecurity defenders specialized Carbon access and offered scientists six months of free access via Google Antigravity. Two researchers we track were already circulating the post by the end of Monday.