# Vals AI Finds Multi-Agent Coding Teams Cost 5x More but Rarely Score Better

AlphaSignal, 2026-10-09. [Read the original](https://alphasignal.ai/news/vals-ai-finds-multi-agent-coding-teams-cost-5x-more-but-rarely-score-better).

Vals AI ran GPT 6 Sol and Claude Opus 5.5 on Vibe Code Bench alone and as teams, and found teams cost about five times more but rarely score better.