Multi-agent workflows may unintentionally increase Codex usage due to fixed overhead costs
ChrisMasterton · ghdev · 2026-08-21
An OpenAI Codex issue highlights that multi-agent workflows can consume more usage than single-agent workflows, even when using smaller models for subtasks.
Root Cause:
- Every spawned subagent incurs a fixed context cost (system instructions, tool schemas, skill metadata, repo context, etc.).
- This fixed overhead is multiplied across multiple agents, potentially outweighing the savings from using smaller models.
- For example, splitting an investigation into 5 agents requires each to load a similar baseline context before doing minimal unique work.
Requested Improvements:
- Display per-agent usage in the UI, distinguishing cached/uncached input and output.
- Visualize context inheritance/forking between parent and child agents.
- Add warnings when spawning agents is likely to cost more than serial execution.
More from coding & agent
- Photon 2.0 launches unified messaging API for AI agents across iMessage and more — Scobleizer · 2026-10-16
- Claude Opus 5.5 effort levels tested: low reasoning iterates 9x faster than max — tom_doerr · 2026-10-16
- Photon 2.0 lets AI agents reach users via WhatsApp, Telegram, SMS and voice in one click — techNmak · 2026-10-16
- Figma launches Guidelines and Code Sources so AI agents can follow your design system — JaynitMakwana · 2026-10-16
- Ex-DeepMind Eng Lead's AI Calling Agent Fo Cut Hang-Up Rate from 80% to 20-30% — julianweisser · 2026-10-16
- SPACE Open-Sources Its Sandbox Platform for Agentic Code Execution — marktenenholtz · 2026-10-16