A 600-run benchmark by Ruby committer Yusuke Endoh tested Claude Code (Opus 4.6) across 13 programming languages by having it implement a simplified Git. Dynamic languages — Ruby ($0.36/run, 73s), Python ($0.38/run, 74s), and JavaScript ($0.39/run, 81s) — were consistently the fastest, cheapest, and most reliable. Statically typed languages cost 1.4–2.6x more, with C being the priciest mainstream option at $0.74/run. Adding type checkers to dynamic languages (mypy for Python, Steep for Ruby) imposed 1.6–3.2x slowdowns, largely attributed to increased thinking-token usage. Only 3 failures occurred across all runs — two in Rust and one in Haskell. The benchmark is limited to ~200-line prototyping-scale code and does not measure code quality, maintainability, or runtime performance. Full dataset and source code are available on GitHub.
Sort: