Agentic leaderboards are shifting from raw SWE-Bench to blended tool-use benchmarks.

June leaderboards increasingly score a weighted blend of Terminal-Bench 2.0, BrowseComp, and OSWorld-Verified instead of SWE-bench alone, with BenchLM weighting "agentic" browse-and-do workflows at 22%, its highest single category. When you evaluate a tool, the headline chat or single-file coding score understates real agent performance. What matters now is how the model executes multi-step, tool-using, environment-acting work end to end. Stop reading the one number on the model card.