AI agent teams waste massive tokens for barely measurable quality gains, research finds
AI Summary
Research from Vals AI reveals that teams of AI agents consume up to 5.1 times more tokens than solo agents while delivering barely measurable quality improvements, with only one of four tests showing gains. Anthropic's data confirms that quality plateaus beyond ten agents even as token costs continue to rise exponentially.
From the source
Teams of AI agents barely outperform solo agents but cost up to 5.1x more, according to Vals AI. Only one out of four tests with GPT-6 Sol and Claude Opus 5.5 showed a measurable gain. Anthropic's own data backs this up: beyond ten agents, quality plateaus while token costs keep climbing. The article AI agent teams waste massive tokens for barely measurable quality gains, research finds appeared first on The Decoder.
The full text couldn't be loaded here (the source may require a subscription).
View original at The Decoder