Google launches Gemini 4 Carbon frontier model in phased rollout
Understand the story
On October 29, 2026, Google DeepMind released Gemini 4 Carbon, positioned as its most powerful frontier model, focusing on cybersecurity and long-horizon agent capabilities. According to evaluation results published by Google DeepMind, the model ranked first with a score of 59 on the Artificial Analysis Intelligence Index v5.0, surpassing the earlier released Argon (The Decoder reported Argon's score as 50); on cybersecurity-related benchmarks, it achieved 88.2% on SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on CWE-bench. The Decoder reported that the model achieved 88.2% on SEC-bench Pro in safety evaluations without added guardrails. Both Ars Technica and The Decoder reported that Gemini 4 Carbon is Google's new frontier model launched two weeks after the release of Gemini 4 Argon, and Google said it reaches new thresholds in coding, agent tasks, and scientific analysis; however, it is currently being rolled out in phases and users cannot yet access it directly. A follow-up report from The Verge added that the model ranks first on the Artificial Analysis Intelligence Index v5.0, ahead of Claude Opus 5.5 and GPT-6 Astra. Current public information mainly comes from Google DeepMind's first-party announcements and media reports; third-party testing and further details remain to be seen.
AI overview of reporting · Updated 28 min前
Reporting timeline
Explore different aspects of the story through its reporting.
- The Verge · AIGoogle launches Gemini 4 Carbon with a ‘significant leap’ in cybersecurity
Google has released its new frontier model Gemini 4 Carbon, claiming significant improvements in coding, agent tasks, and cybersecurity. It currently ranks first on the Artificial Analysis Intelligence Index (v5.0), ahead of Claude Opus 5.5 and GPT-6 Astra.
- The DecoderSelectedGoogle Deepmind releases Gemini 4 Carbon just two weeks after Argon with gains across the board
Google DeepMind launched Gemini 4 Carbon just two weeks after releasing Argon, topping the Artificial Analysis Intelligence Index v5.0 with a score of 59 (versus 50 for Argon), and scoring 88.2% on SEC-bench Pro in safety evaluations without guardrails.
- Ars Technica · AIGoogle announces Gemini 4 Carbon, its latest frontier AI model that you can't use
Two weeks after releasing Gemini 4 Argon, Google announced its new frontier model Gemini 4 Carbon, claiming it reaches a new threshold in coding, agentic tasks, and scientific analysis. It is being rolled out in phases, and users cannot access it directly yet.
- Google DeepMindSelectedGemini 4 Carbon: Our next frontier model
Google DeepMind released Gemini 4 Carbon, positioned as its strongest model, ranking first on the Artificial Analysis Intelligence Index v5.0 with a score of 59, and scoring 88.2% on the cybersecurity benchmark SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on CWE-bench.
Story heat trend
Current heat19·Comparable peak19（10/29 04:00）·Comparable change over 24 hours –
The trend compares the same continuously observed sources, which may cover fewer sources than the current heat score. Hover or select the chart to inspect each hour; use the left and right arrow keys to navigate.