Multi-Agent Teams Cost 5x More for Almost No Benchmark Gain, Vals AI Study Finds
A Vals AI study on GPT-6 Sol and Claude Opus 5.5 found that orchestrating teams of agents cost five times more while producing a significant improvement in just one of four comparisons.
The multi-agent hype is getting an empirical cold shower. A Vals AI study, reported by @FadyEid, found that AI agent teams cost roughly five times more than single-agent approaches while delivering a statistically significant performance gain in only one of four comparisons tested. The models under evaluation were current flagships — GPT-6 Sol and Claude Opus 5.5 — so this isn't a case of weak models failing to coordinate.
Unlock the full briefing
Get every story in today's briefing, the full archive, and the daily AI intelligence brief.
All stories today
Full archive
Daily brief
Cancel anytime. Payments powered by Stripe.