ToolOrchestra: Elevating Intelligence via Efficient Model and Tool Orchestration
This addresses the challenge of computational expense and performance in tool-augmented reasoning for AI systems, offering a scalable solution, though it appears incremental as it builds on existing tool-use agents.
The paper tackles the problem of solving complex tasks like the Humanity's Last Exam (HLE) by introducing ToolOrchestra, a method for training small orchestrators to coordinate models and tools, resulting in an 8B model that achieves higher accuracy at lower cost, such as 37.1% on HLE outperforming GPT-5 with 2.5x efficiency.
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.