Coding-agent benchmarks with cost and runtime

Artificial Analysis now compares coding agents across task benchmarks while also reporting cost per task, token use, and execution time. Why it matters: Agent choice should not be based only on score. Cost, latency, and task type determine whether an agent is practical for repo Q&A, terminal work, or patch generation.

Try this: Use the index as a routing checklist: compare at least score, price, tokens, and wall time before standardizing on a coding agent for a task class.