Sakana AI Taps NVIDIA for Open Model Push

Sakana AI is teaming up with NVIDIA to enhance its approach to collective intelligence, integrating the chip giant's open models into its multi-agent orchestration system. This Sakana AI NVIDIA partnership aims to foster open model innovation from Japan for a global audience.

The core of this collaboration is Sakana Fugu, an intelligence orchestrator that dynamically selects and combines multiple specialized AI models. Behind a single interface, Fugu leverages the strengths of various agents for each task, synthesizing them into a unified response. This modular design allows for continuous improvement as new models become available.

Collective Intelligence Meets Accelerated Computing

NVIDIA's open models, including the NVIDIA Nemotron open model family, will be integrated as specialized agents within Sakana Fugu. These models, featuring open weights and tooling, bring distinct strengths in coding, tool calling, and instruction following to Fugu's agent pool.

This integration moves beyond single model scaling, focusing instead on orchestrating diverse, specialized open models. The goal is to demonstrate how these models become more powerful when coordinated within agentic systems, rather than used in isolation.

Sakana Fugu itself is a language model designed to call and coordinate various LLMs, including instances of itself recursively, to tackle complex, multi-step tasks.

"We are excited to integrate open models, such as NVIDIA’s Nemotron, into Fugu," stated David Ha, Co-Founder and CEO of Sakana AI. "The collaboration is designed to create a reinforcing cycle between open models, orchestration, and real-world use."

Kari Briski, Vice President of Generative AI at NVIDIA, added, "Open models give countries the ability to build AI that reflects their own language, culture and policies. By bringing NVIDIA Nemotron into Sakana Fugu, we're helping developers build more capable AI systems and demonstrating what's possible when open and closed models are intelligently orchestrated."

A Modular, Open Future for AI

The partnership aims to create a feedback loop where real-world usage signals improve both the models and the orchestration layer. This approach reflects a broader industry shift where progress increasingly depends on effectively evaluating, combining, and deploying diverse open models.

As open models proliferate, the ability to orchestrate them becomes critical for building practical and reliable AI systems. NVIDIA's contribution includes its open model stack, support for open model development, and its accelerated computing platform, crucial for scaling this orchestration layer.

This initiative underscores the growing importance of open AI ecosystems, offering transparency and freedom from single-provider lock-in. Sakana AI, from Japan, contributes its unique collective intelligence approach to practical open model utilization.

The collaboration will involve ongoing technical work to observe and enhance Nemotron's performance within Fugu's multi-agent workflows. This includes NVIDIA providing technical guidance on Nemotron integration and evaluation best practices.

Together, Sakana AI and NVIDIA are working towards a new generation of AI systems that are modular, collaborative, and open by design. This AI model orchestration strategy emphasizes that the most capable AI will emerge from coordinated models, not isolated ones, shaping a more open future for AI, originating from Japan for the world.