Many hands might make light work. Many AI agents apparently make accounts payable.
Vals AI tested GPT-6 Sol and Claude Opus 5.5 solo versus in teams across 50 app-building tasks. Every team cost more. Only one of four comparisons showed a statistically significant score improvement.
Opus at maximum effort? Roughly five times the cost, more than twice the time, and no statistically significant score gain.
Running a swarm doesn’t automatically make you AI native. You might just be spinning your wheels and blowing money.
Study by Vals AI, “Do Agent Teams Pay Off?”: https://www.vals.ai/blogs/multi-agent-vibe-code-bench