Lyptus Research finds GPT-5.5 achieved a 12-hour cybersecurity task-solving horizon under a 50-million-token budget

— GPT-5.5 resolved 92.4% of the cyber benchmark tasks.

Original post

OPLA#980Lisan al Gaib|@SCALING01

"GPT-5.5 now saturates our dataset"

8:56 AM · May 27, 2026 View on X