Google unveils Gemini 4 Carbon frontier model with enhanced cybersecurity focus
In short: Google has announced Gemini 4 Carbon, a new high-performance model designed for complex engineering and cybersecurity. The model excels in vulnerability detection, secure patching, and long-horizon reasoning, while introducing improved alignment and prompt-injection resistance.
This summary was generated automatically by AI from Google DeepMind's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1SEC-bench Pro vulnerability detection at 88.2%
- 2ExploitGym success rate of 54.8%
- 3CWE-bench patching performance of 71.0%
- 4Indirect Prompt Injection (IPI) attack success rate reduced to 0.0%
- 5Pricing set at $2/$10 per million tokens for input/output
| Parameter | Before | Now |
|---|---|---|
| Input price | Not specified | $2 per million tokens |
| Output price | Not specified | $10 per million tokens |
| Cached input price | Not specified | 95% off input price |
Why it matters
The model is specifically tuned for agentic workflows that require persistence through multi-step engineering tasks. For teams handling secure software development or automated threat analysis, Carbon offers advanced capabilities in root-causing and patching vulnerabilities.
What it means for AI agents and contact centers
When evaluating Carbon, focus on multi-step reasoning reliability in complex codebases and the effectiveness of its security-oriented prompt-injection defenses. Benchmark it against current agents for your long-running software engineering automations to see if the reduction in fabrication and overreach warrants migration.
🧪 Worth Testing
The model's significant improvements in vulnerability detection and secure code generation are highly relevant for automating internal secure development lifecycles.
Sources
- Google DeepMindOfficialPrimary sourceOriginal article →„Gemini 4 Carbon: Our next frontier model“28 Oct 2026, 19:00
- Published by source
- 28 Oct 2026, 19:00
- Found by our system
- 28 Oct 2026, 19:22
- Summary generated
- 28 Oct 2026, 19:22
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy