Scaling Small Agents Through Strategy Auctions

TL;DR

This paper investigates how task complexity mediates the performance gap between small and large language-model agents and introduces Strategy Auctions for Workload Efficiency (SALE), a test-time, marketplace-inspired routing framework. SALE enables heterogeneous agents to bid for tasks using short strategic plans, evaluated by a cost–value mechanism and refined through shared auction memory, effectively reallocating work to smaller, cheaper models without a separate router. Across deep search and coding tasks, SALE reduces reliance on the largest model by about 53% and lowers total cost by 35%, while achieving or exceeding the largest agent’s pass@1, with negligible overhead. The work argues for a systems-level rethinking of agentic AI, where coordination, memory, and market-inspired dynamics can outperform pure scaling of individual models, potentially democratizing access to capable AI while highlighting trade-offs and environmental considerations.

Abstract

Small language models are increasingly viewed as a promising, cost-effective approach to agentic AI, with proponents claiming they are sufficiently capable for agentic workflows. However, while smaller agents can closely match larger ones on simple tasks, it remains unclear how their performance scales with task complexity, when large models become necessary, and how to better leverage small agents for long-horizon workloads. In this work, we empirically show that small agents' performance fails to scale with task complexity on deep search and coding tasks, and we introduce Strategy Auctions for Workload Efficiency (SALE), an agent framework inspired by freelancer marketplaces. In SALE, agents bid with short strategic plans, which are scored by a systematic cost-value mechanism and refined via a shared auction memory, enabling per-task routing and continual self-improvement without training a separate router or running all models to completion. Across deep search and coding tasks of varying complexity, SALE reduces reliance on the largest agent by 53%, lowers overall cost by 35%, and consistently improves upon the largest agent's pass@1 with only a negligible overhead beyond executing the final trace. In contrast, established routers that rely on task descriptions either underperform the largest agent or fail to reduce cost -- often both -- underscoring their poor fit for agentic workflows. These results suggest that while small agents may be insufficient for complex workloads, they can be effectively "scaled up" through coordinated task allocation and test-time self-improvement. More broadly, they motivate a systems-level view of agentic AI in which performance gains come less from ever-larger individual models and more from market-inspired coordination mechanisms that organize heterogeneous agents into efficient, adaptive ecosystems.

Figures (9)

• Figure 1: Pass@1 accuracy on deep search and coding tasks (a) and average trace length in million tokens (b). We show the effective price per million tokens $\pi(a_d)$ for Qwen3 agents from smallest to largest ($d =$ {4B, 8B, 14B, 32B}).
• Figure 2: An illustration of the sale pipeline. Given a task $t$, each agent $a_i$ proposes a strategic plan $s_{t,i}$ as its bid. Bids are evaluated by cost $C_{t,i}$ and value $V_{t,i}$, and a provisional winner is selected by minimizing cost-minus-value. Agents cheaper than the provisional winner may then refine their strategies using similar past successes and failures retrieved from the auction memory, after which a final winner is selected and its strategy is executed.
• Figure 3: Performance–cost trade-offs for deep search (top row) and coding (bottom row) across task-complexity bins. At a given price per million tokens $\pi$, the sale auction ensemble consistently attains substantially higher pass@1 than would be predicted by the approximate linear scaling trend observed for individual Qwen3 agents, showing that it systematically exceeds the expected performance–cost curve.
• Figure 4: sale's average workload allocation across the 4B, 8B, 14B, and 32B agents for deep search (top) and coding (bottom) tasks, stratified by task complexity $\tau(t)$. Bar labels indicate the share of all tasks assigned to each agent.
• Figure 5: Cumulative share of tasks routed to the smallest agent over time. Solid lines show the mean across 5 runs with randomized task orderings; shading denote ±1 standard deviation. An upward trend indicates that the local selection rate exceeds the historical average, reflecting increased delegation to the smallest agent as auction history accumulates.