Google & MIT Study: Multi-Agent Setups Underperform Single Agents Above 45% Accuracy
pliang279 · x · 2026-08-14
A joint study by Google Research, DeepMind, and MIT tested 260 agent configurations across six benchmarks, holding tools, prompts, and compute fixed to isolate the impact of topology.
The research reveals a counterintuitive rule: once a single agent scores above 45% on a task, adding more agents makes performance worse.
- Error Amplification: Single agent (1.0x) drastically outperforms decentralized (7.8x) and independent topologies (17.2x).
- Efficiency: Success per thousand tokens is 67.7 for a single agent, compared to just 13.6 for hybrid setups.
- SWE-bench: Every multi-agent variant scored worse than a single agent.
This indicates that coordination costs in multi-agent systems quickly overtake the remaining upside.
Related event: Studies Reveal Hidden Costs of Multi-Agent Collaboration(2 posts)→
More from coding & agent
- How do you debug bad agent runs when you already have tracing? — Sensitive-Parsnip-12 · 2026-10-16
- Where does pre-task context live for agents? Devs debate workflows beyond handoff — shishen2026 · 2026-10-16
- No-instruction autonomous agent wakes up, sets own goals, and tries hacking by default — sethlazar · 2026-10-16
- CODESTRUCT reframes codebases as AST action spaces to fix brittle LLM code agents — burkov · 2026-10-16
- Would you install another SDK for agent debugging on top of your tracing stack? — Sensitive-Parsnip-12 · 2026-10-16
- Law student builds all-in-one AI study platform LexLab Studio with ChatGPT and Muse — Acrobatic_Rhubarb598 · 2026-10-16