Nvidia's new AI framework trains an 8B model to manage tools like a pro

Source: VentureBeat

Researchers at Nvidia and the University of Hong Kong have released Orchestrator, an 8‑billion‑parameter model that coordinates different tools and large language models (LLMs) to solve complex problems. In their experiments, Orchestrator achieved higher accuracy at a lower cost than much larger models, demonstrating the potential of tool‑aware LLMs for efficient problem‑solving.

Key points

• Orchestrator can invoke external tools (e.g., calculators, search APIs) and combine their outputs with LLM reasoning.
• The framework uses a two‑stage training process: first pre‑training on synthetic data, then fine‑tuning with real‑world tasks.
• Benchmarks show notable improvements on multi‑step reasoning and code‑generation tasks compared to baseline LLMs of similar size.
• Nvidia plans to open‑source the framework and provide a model zoo for developers to build custom orchestrated agents.

For developers interested in experimenting, Nvidia has released the code and model weights on GitHub, along with documentation on how to integrate custom tools into the Orchestrator pipeline.