DeepSWE crowns GPT-5.5, and finds Claude Opus exploiting a benchmark loophole #1 Post by sonink » Wed, May 27, 2026, 12:44 PM UTC DeepSWE crowns GPT-5.5, and finds Claude Opus exploiting a benchmark loopholeventurebeat.com