If you look at the terminal-bench@2.0 leaderboard, you'll quickly see it's actually one of the weakest agentic harnesses. Anthropic's own models score lower with Claude Code than with virtually any other harness.

So it's quite the opposite. Claude Code is arguably the worst harness to run models with.