Google announces Gemini 4 Carbon for coding and cyber defense
Google DeepMind
- Models
- Benchmarks
- Security
- #gemini
- #carbon
- #artificial analysis
Google DeepMind announced Gemini 4 Carbon, describing it as its next frontier model and strongest to date, with state-of-the-art cybersecurity and long-horizon capabilities. According to the announcement, Carbon ranks first on v5.0 of the Artificial Analysis Intelligence Index at 59 and leads benchmarks for finding vulnerabilities, exploitation and patching. It follows Gemini 4 Argon, launched two weeks earlier.
What changed
Carbon is positioned as a step beyond Gemini 4 Argon, with reported gains in coding, agentic tasks and cybersecurity defense. The announcement cites SEC-bench Pro at 88.2% for vulnerability finding, ExploitGym at 54.8% and CWE-bench at 71.0%, and says internal tests show reduced misalignment and fabrication rates.
Who it helps
Developers, security teams and enterprises evaluating frontier AI models, plus observers tracking competition among major AI labs.
Context and limitations
- Publisher claims were not independently tested. The excerpt does not establish public availability or access terms.
- Based on a short source excerpt; the full article was not analyzed.