AI agent teams waste massive tokens for barely measurable quality gains, research finds
decoder ·
Recent research indicates that AI agent teams often consume excessive computational resources while delivering negligible improvements in output quality. Vals AI evaluated GPT-6 Sol and Claude Opus 5.5 on the Vibe Code Bench, finding that multi-agent setups cost between 1.8x and 5.1x more than single agents with minimal performance gains. Only one comparison showed a statistically significant improvement, suggesting that the added expense is rarely justified. Internal tests by Anthropic further demonstrated that scaling from ten to 100 agents yielded only slight score increases despite higher token usage. OpenAI researcher Noam Brown confirmed that multi-agent systems primarily accelerate task completion rather than enhancing quality, noting that parallelization benefits vary significantly by task type. Experts warn against using large agent swarms due to coordination breakdowns and high costs, describing it as a wasted investment for most applications.
decoder