# The Coordination Tax and the Heterogeneity Dividend: Break-Even Conditions for Multi-Agent LLM Pipelines

DevFeed: [The Coordination Tax and the Heterogeneity Dividend: Break-Even Conditions for Multi-Agent LLM Pipelines](<https://devfeed.tech/articles/the-coordination-tax-and-the-heterogeneity-dividend-break-even-conditions-for-multi-agent-llm-pipelines-107951.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.37700>)

Author: Andreas Zettl, Diego Gosmar

Published: 2026-10-30T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [agent to agent communication](<https://devfeed.tech/topics/agent-to-agent-communication.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>)

Tags: [a2a](<https://devfeed.tech/tags/a2a.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [agent](<https://devfeed.tech/tags/agent.md>), [call](<https://devfeed.tech/tags/call.md>), [choice](<https://devfeed.tech/tags/choice.md>), [claude](<https://devfeed.tech/tags/claude.md>), [claude-opus-5-5](<https://devfeed.tech/tags/claude-opus-5-5.md>), [cost](<https://devfeed.tech/tags/cost.md>), [costs](<https://devfeed.tech/tags/costs.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-ma](<https://devfeed.tech/tags/cs-ma.md>), [dividend](<https://devfeed.tech/tags/dividend.md>), [forecast](<https://devfeed.tech/tags/forecast.md>), [haiku](<https://devfeed.tech/tags/haiku.md>), [half](<https://devfeed.tech/tags/half.md>), [humaneval](<https://devfeed.tech/tags/humaneval.md>), [llm](<https://devfeed.tech/tags/llm.md>), [model](<https://devfeed.tech/tags/model.md>), [moving](<https://devfeed.tech/tags/moving.md>), [multi-agent](<https://devfeed.tech/tags/multi-agent.md>), [pipelines](<https://devfeed.tech/tags/pipelines.md>), [tax](<https://devfeed.tech/tags/tax.md>)

## AI overview

The study models when a multi-agent pipeline's coordination tax--decomposition overhead plus protocol handoffs--can be offset by routing roles to cheaper models, and tests the model with OFP on 135 HumanEval and ClassEval tasks across 2,430 runs. OFP adds 7.2 ms per handoff (0.11% of pipeline time). On these tasks, a planner-coder-reviewer pipeline using the strong model costs 3.22x as much and takes 3.24x as long as one strong call without solving more tasks. Assigning planner and reviewer to the cheaper model cuts pipeline cost to 0.39x, but pass@1 falls by 6.7 points; the pipeline remains slower, less accurate, and dearer than the strong call, partly because cheap roles inflate token volume and the break-even forecast misses by 19%. An oracle choosing between the two single-model calls solves 94.3% of runs at half the strong call's cost per solved task. These findings are limited to the reported benchmarks and models.

## Source excerpt

arXiv:2610.37700v1 Announce Type: cross Abstract: Multi-agent LLM pipelines built on interoperability protocols such as the Open Floor Protocol (OFP) or A2A are usually justified on capability grounds, with their runtime cost treated as an overhead of unknown size. Comparing such a pipeline with one model call is underspecified, because a pipeline can give each role a different model. We give a closed-form model that sets the coordination tax (decomposition overhead plus per-handoff protocol cost) against the heterogeneity dividend from routing work to a cheaper model, with a break-even threshold and a feasibility bound, $\alpha < 1/r$, and measure every term, with OFP as the protocol under test, on 135 HumanEval and ClassEval tasks in 2430 runs of six architectures with Claude Opus 5.5 and Haiku 5.5. The protocol adds 7.2 ms per handoff, 0.11% of pipeline wall time. Decomposition does not pay on these tasks: a planner-coder-reviewer pipeline costs 3.22x as much as a single strong call, takes 3.24x as long, and solves no more tasks. Moving planner and reviewer to the cheap model cuts the pipeline's cost to 0.39x but changes pass@1 by -6.7 points, leaving it dearer, slower and less accurate than the single call, and the break-even forecast misses by -19% because the cheap roles inflate token volume. The saving lies in choosing a model per task instead: an oracle choice between the two single calls solves 94.3% of runs at half the strong call's cost per solved task.