Claude Opus 4.6 Shatters Records — #1 on SWE-Bench with 72% Autonomous Resolve Rate
Anthropic's Claude Opus 4.6 achieves state-of-the-art across SWE-Bench, GPQA, and MATH, with breakthrough agentic coding that resolves real GitHub issues autonomously.