OpenAI GPT-5.6 Sol Achieves State-of-the-Art in Cyber Benchmarks

GPT-5.6 Sol outperformed Mythos 5 on UK AISI's narrow cyber tasks and "The Last Ones" benchmarks, fully solving the latter in 7 out of 10 attempts. On the CyberGym ExploitGym benchmark, GPT-5.6 Sol scored 293/869 in a 6-hour window, surpassing Mythos Preview's 204 under the same conditions.