Sakana AI adds Nvidia Nemotron to Fugu orchestrator to rival frontier models

Tokyo-based Sakana AI expanded its Fugu orchestrator to include Nvidia's open Nemotron models. Fugu is itself a language model trained to call other LLMs from an agent pool, dynamically picking which to combine, delegating subtasks, and synthesizing results behind a single API, with Nemotron acting as a specialist in programming, tool use, and instruction-following. Sakana argues the most powerful AI emerges from many models working together, and that the modular design reduces dependence on any single vendor. It claims its Fugu Ultra variant performs on par with Anthropic's Fable 5 in its own benchmarks.

View full digest for July 17, 2026