ToolOrchestra: Elevating Intelligence via Efficient Model and Tool Orchestration

TL;DR

ToolOrchestra introduces a small, reinforcement learning-based orchestrator that manages multiple tools to enhance the accuracy and efficiency of large language models on complex tasks, outperforming larger models like GPT-5.

Contribution

The paper presents a novel training method for small orchestrators that coordinate tools, achieving superior performance and efficiency compared to existing tool-use agents and large models.

Findings

Orchestrator achieves higher accuracy at lower cost than previous agents.

On HLE, Orchestrator outperforms GPT-5 with 2.5x efficiency gain.

Generalizes robustly to unseen tools, maintaining performance.

Abstract

Large language models are powerful generalists, yet solving deep and complex problems such as those of the Humanity's Last Exam (HLE) remains both conceptually challenging and computationally expensive. We show that small orchestrators managing other models and a variety of tools can both push the upper bound of intelligence and improve efficiency in solving difficult agentic tasks. We introduce ToolOrchestra, a method for training small orchestrators that coordinate intelligent tools. ToolOrchestra explicitly uses reinforcement learning with outcome-, efficiency-, and user-preference-aware rewards. Using ToolOrchestra, we produce Orchestrator, an 8B model that achieves higher accuracy at lower cost than previous tool-use agents while aligning with user preferences on which tools are to be used for a given query. On HLE, Orchestrator achieves a score of 37.1%, outperforming GPT-5 (35.1%) while being 2.5x more efficient. On tau2-Bench and FRAMES, Orchestrator surpasses GPT-5 by a wide margin while using only about 30% of the cost. Extensive analysis shows that Orchestrator achieves the best trade-off between performance and cost under multiple metrics, and generalizes robustly to unseen tools. These results demonstrate that composing diverse tools with a lightweight orchestration model is both more efficient and more effective than existing methods, paving the way for practical and scalable tool-augmented reasoning systems.