Google DeepMind Launches Carbon Model, Tops AI Intelligence Benchmarks

Google DeepMind has launched the Carbon model, surpassing the previous Argon model with impressive results on AI intelligence benchmarks. Carbon boasts a success rate of 88.2% in vulnerability detection, 54.8% in exploitation, and 71.0% in patching, establishing it as a leading model in cybersecurity. Initially, Carbon will be available to users of the Fairwind Program, with a promise of broader developer access coming soon. This move is significant for enhancing AI capabilities in long-horizon tasks.