Sundar Pichai unveils Gemini 4 Carbon with record security scores
Google CEO Sundar Pichai just introduced Gemini 4 Carbon, the newest and most powerful model in the Gemini AI lineup.
This version is a big leap for cybersecurity and coding: it scored a record 59 on the newly updated v5.0 of the Artificial Analysis Intelligence Index, and Google says it outperforms rivals like GPT-6 Astra and Claude Opus 5.5 across multiple benchmarks, and nailed an 88.2% on a key security test for catching software bugs.
Gemini 4 carbon halves misinformation rates
Gemini 4 Carbon also cuts down on false information by one-half and slashes error alerts by almost 80%.
It's especially good at fixing code safely, even completely blocking all simulated hacking attempts during strict safety evaluations.
Starting this week, a select group of trusted cybersecurity defenders will get early access. After that, Google will slowly expand access to more partners through its Fairwind Program so they can help fine-tune things before a wider release.