Log-scale replot of OpenAI swarm data: 4x agents cost ~2x total tokens

tobyordoxford · x · 2026-09-22

tobyordoxford re-plotted OpenAI's multi-agent swarm graphs on a logarithmic x-axis, showing all swarm sizes scale logarithmically with the same slope: every 4x increase in agents takes about 2x total tokens. From the original charts, a 1-agent swarm needs 50% of the tokens of a 4-agent swarm and 25% of a 16-agent swarm to reach each capability level — meaning solo agents remain markedly more compute-efficient than swarms.

Related event: Oxford's Toby Ord Quantifies Swarm Scaling: Agent Clusters Far Less Efficient Than Longer Chains of Thought(10 posts)→

Original post →

More from coding & agent

coding & agent channel →