I pitted Claude Opus 5 against Kimi K3 with 15 demanding coding and infrastructure prompts to see which AI model actually ...