Nemotron meets Fugu: Sakana AI's bet that open models win as a swarm, not alone

Sakana AI announced a collaboration with NVIDIA to bring the Nemotron family of open-weight models into Sakana Fugu, the company's multi-agent orchestration system. The partnership tests a central thesis: that coordinating many open, specialized models can produce results comparable to monolithic frontier systems while remaining modular, adaptable, and independent of any single provider. This isn't a new debate, the open-model ecosystem has long argued that local, specialized models can handle real tasks without the overhead of giant black boxes.

Fugu works as an intelligence orchestrator. Behind a single interface and API, it dynamically selects, coordinates, and combines the strengths of multiple underlying models and agents, choosing the right capability for each task and synthesizing them into a single response. Because new and better models can be added over time, the system improves not only through its own development but also through the progress of the broader AI ecosystem. It is never tied to the strengths, limitations or availability of any single model. That philosophy echoes a wider shift: Sakana has long argued that model choice should be the router's problem, not the developer's.

Nemotron, which NVIDIA describes as an open model family with open weights and tooling, will join Fugu's agent pool as specialized models. The models have demonstrated strength in coding, tool calling, and instruction following. The collaboration is designed to create a reinforcing cycle: Fugu gains access to a deeper pool of specialized capabilities, while NVIDIA can evaluate how Nemotron performs when coordinated as part of agentic, multi-step workflows. NVIDIA's own benchmarks suggest that better embeddings can pay for themselves in reduced agent runtime, which fits neatly into the orchestrated cost calculus.

"We're excited to collaborate with NVIDIA to build the next generation of Fugu orchestration models together, by incorporating leading open-weights models like Nemotron," said David Ha, co-founder and CEO of Sakana AI.

Kari Briski, vice president of generative AI at NVIDIA, said that open models give countries "the ability to build AI that reflects their own language, culture and policies" and that the collaboration demonstrates "what's possible when open and closed models are intelligently orchestrated."

The next phase involves concrete technical work. Once Nemotron is integrated as a specialized agent in an upcoming version of Fugu, the teams will collaborate to continuously observe and improve Nemotron's performance within the orchestration system. NVIDIA will provide technical guidance on Nemotron recipes and evaluation best practices. The companies will evaluate how Nemotron performs within Fugu's multi-agent workflows and use those findings to inform future improvements to both Nemotron and Fugu. This feedback loop mirrors what training-trick research has shown: that exposing models to real operational noise can dramatically improve their production stability.

The collaboration also carries a broader message for the open model ecosystem. Sakana AI's approach to collective intelligence, which the company has explored in domains ranging from evolutionary creativity to self-recognizing modular hardware, translates here into a system that treats each open model as a specialized agent rather than a standalone substitute for a frontier model. The bet is that the orchestration layer itself becomes a scaling path, and that real-world usage can feed back into improving both the models and the coordination layer.

"No single model is likely to hold every advantage across every task, language, modality and enterprise environment," the announcement states. "That makes orchestration a critical layer for the next stage of open AI, turning a diverse model ecosystem into practical, reliable systems."