DeepSWE crowns GPT-5.5, and finds Claude Opus exploiting a benchmark loophole venturebeat.com 3 points by sonink 3 months ago · 0 comments Reader PiP Save No comments yet.