Sakana Fugu

What is Sakana Fugu?

Sakana Fugu is a revolutionary multi-agent system that dynamically orchestrates a diverse pool of specialized AI models through a single, OpenAI-compatible API. Instead of relying on predefined workflows or human-designed agent roles, Fugu learns to assemble and coordinate expert agents on the fly, discovering non-obvious collaboration patterns that yield superior performance on complex, multi-step tasks. This approach enables frontier-level performance without single-vendor dependency, as Fugu can seamlessly switch between top models from different providers to optimize for accuracy, cost, and speed.

Key features include dynamic model orchestration, automatic model selection and switching, and flexible agent pool control. Users can opt out of specific providers or models to meet data, privacy, compliance, or organizational requirements. Fugu is built for coding, reasoning, and other quality-critical workflows, where it coordinates expert agents to produce stronger, more reliable results. The system is grounded in two ICLR 2026 papers—TRINITY and the Conductor—which demonstrate how systems can learn to assemble, route, and coordinate expert agents for each task.

Use cases include automated code generation, complex reasoning tasks, data analysis, and any workflow requiring high-quality outputs from multiple specialized models. By reducing API complexity and improving cost-performance, Fugu allows developers and enterprises to plug collective intelligence directly into their workflows with minimal integration effort. Note that Fugu is not yet available in the EU/EEA while the team works toward compliance with GDPR and EU-specific regulations. With its innovative learned orchestration, Sakana Fugu represents a significant leap forward in multi-agent AI systems, offering a powerful, flexible, and efficient solution for tackling the most demanding AI tasks.

Key Features

dynamic model orchestration automatic model selection flexible agent pool control provider opt out multi step task coordination frontier level performance cost optimization speed optimization non obvious collaboration patterns OpenAI compatible API

Who is it for?

AI engineers, software developers, data scientists, research teams, enterprise IT, compliance officers, platform teams