Sakana AI and NVIDIA announced a collaboration on July 15, 2026, to integrate NVIDIA's open model stack, specifically the Nemotron model family, into Sakana Fugu, a multi-agent orchestration system. This partnership aims to enhance Sakana Fugu's ability to coordinate multiple specialized models, allowing it to dynamically select and combine strengths from various agents for complex, multi-step tasks. By incorporating Nemotron, which offers specialized capabilities in coding, tool calling, and instruction following, Sakana AI seeks to demonstrate that open models become more effective when orchestrated within a modular agentic architecture rather than operated in isolation.

At the core of this initiative is Sakana Fugu, an intelligence orchestrator that operates behind a single interface and API to synthesize responses from diverse models. This design allows for a modular, adaptable system that can integrate new, improved models over time while reducing dependency on any single provider or model limitations. Early evaluations have shown that this orchestration-based approach performs competitively with frontier AI systems, suggesting that model coordination represents a viable scaling path for future development.

As part of the technical integration, the teams will observe and optimize how Nemotron performs within Fugu's multi-agent workflows. NVIDIA will provide technical guidance on model recipes and evaluation best practices, creating a feedback loop where real-world usage signals help improve both the underlying models and the orchestration layer. This effort is intended to benefit developers and enterprises by offering increased flexibility, choice, and modularity in AI deployments.

Both organizations emphasize that this collaboration supports the broader growth of open model ecosystems, which are increasingly seen as foundational to worldwide AI innovation. By combining Sakana AI's focus on collective intelligence—a design philosophy drawing inspiration from natural swarm behavior—with NVIDIA’s accelerated computing platform, the companies aim to build more resilient AI systems. The project is positioned to contribute to the global landscape of open AI while advancing research and practical applications originating from Japan.

Sakana AI and NVIDIA announced a collaboration on July 15, 2026, to integrate NVIDIA's open model stack, specifically the Nemotron model family, into Sakana Fugu, a multi-agent orchestration system. This partnership aims to enhance Sakana Fugu's ability to coordinate multiple specialized models, allowing it to dynamically select and combine strengths from various agents for complex, multi-step tasks. By incorporating Nemotron, which offers specialized capabilities in coding, tool calling, and instruction following, Sakana AI seeks to demonstrate that open models become more effective when orchestrated within a modular agentic architecture rather than operated in isolation.

At the core of this initiative is Sakana Fugu, an intelligence orchestrator that operates behind a single interface and API to synthesize responses from diverse models. This design allows for a modular, adaptable system that can integrate new, improved models over time while reducing dependency on any single provider or model limitations. Early evaluations have shown that this orchestration-based approach performs competitively with frontier AI systems, suggesting that model coordination represents a viable scaling path for future development.

As part of the technical integration, the teams will observe and optimize how Nemotron performs within Fugu's multi-agent workflows. NVIDIA will provide technical guidance on model recipes and evaluation best practices, creating a feedback loop where real-world usage signals help improve both the underlying models and the orchestration layer. This effort is intended to benefit developers and enterprises by offering increased flexibility, choice, and modularity in AI deployments.

Both organizations emphasize that this collaboration supports the broader growth of open model ecosystems, which are increasingly seen as foundational to worldwide AI innovation. By combining Sakana AI's focus on collective intelligence—a design philosophy drawing inspiration from natural swarm behavior—with NVIDIA’s accelerated computing platform, the companies aim to build more resilient AI systems. The project is positioned to contribute to the global landscape of open AI while advancing research and practical applications originating from Japan.