Google Launches Gemini 4 Carbon, Claims World's Most Powerful AI Title with Cybersecurity Focus

Google DeepMind unveiled Gemini 4 Carbon on October 28, claiming the title of the world's most powerful AI model. The model scored 59 on the Artificial Analysis Intelligence Index, placing it first ahead of Anthropic's Claude Opus 5.5 at 58 and OpenAI's GPT-6 Astra at 57. Elon Musk called the model "impressive," and Polymarket put the odds of Google leading the AI race at year-end at 41% following the announcement. Google stock edged higher on the news.

Cybersecurity is Gemini 4 Carbon's defining capability. On SEC-bench Pro, the model identified real software vulnerabilities 88.2% of the time. On ExploitGym, which tests whether flaws can be weaponized, it scored 54.8%, well ahead of GPT-6 Astra at 42.4% and Claude Opus 5.5 at 34.5%. On CWE-bench, which measures the ability to write working patches, it scored 71.0%. Google's Chrome team reported that Carbon produced patches more than ten times better than the next best model from a rival lab. Google also says the model generates secure code at more than twice the success rate of its predecessor, Argon.

Beyond security, Google says Carbon can sustain complex coding work for days without getting stuck, making it suited to long-horizon engineering tasks. The model is priced at $2 per million input tokens and $10 per million output tokens. Because of its ability to find and exploit vulnerabilities, Google is restricting initial access to a limited group of vetted cyber defenders before a broader rollout. Google also simultaneously launched Xreal Aura, Android XR-powered glasses developed with Xreal, priced from $1,099 and available to order in the US, Canada, and Korea.