Sakana Fugu: The Multi-Agent AI System That Works Like a Team

Sakana Fugu：像团队一样协作的多智能体 AI 系统

We’ve all been there: copy-pasting a prompt from ChatGPT to Claude, and then to Gemini, trying to find which AI gives the best answer. Different AI models have different strengths. Some are excellent programmers, some write beautifully, and others are master logicians. But constantly switching between them is slow and frustrating. 我们都有过这样的经历：在 ChatGPT、Claude 和 Gemini 之间反复复制粘贴提示词，试图找出哪一个 AI 能给出最好的答案。不同的 AI 模型各有千秋：有的擅长编程，有的文笔优美，有的则是逻辑大师。但不断在它们之间切换既缓慢又令人沮丧。

Sakana AI has introduced a brilliant solution: Sakana Fugu. Fugu is a multi-agent AI system that bundles multiple frontier models into a single, seamless package. To you, it looks like a single chatbot. But behind the scenes, it acts as a project manager, coordinating a team of top-tier AI models to solve your task. Sakana AI 推出了一项绝妙的解决方案：Sakana Fugu。Fugu 是一个多智能体 AI 系统，它将多个前沿模型整合为一个无缝的整体。对用户而言，它看起来就像一个单一的聊天机器人；但在幕后，它扮演着项目经理的角色，协调一支顶尖 AI 模型团队来完成你的任务。

What is Sakana Fugu?

什么是 Sakana Fugu？

Sakana Fugu is a collaborative AI framework. Instead of relying on one massive AI model to do all the work, Fugu orchestrates a pool of specialized models (such as GPT-5, Claude Opus, and Gemini Pro). When you give Fugu a complex, multi-step prompt, it: Sakana Fugu 是一个协作式 AI 框架。它不再依赖单一的大型 AI 模型来完成所有工作，而是调度一个由专业模型（如 GPT-5、Claude Opus 和 Gemini Pro）组成的资源池。当你给 Fugu 一个复杂的多步骤提示词时，它会：

• Analyzes the task and breaks it down into steps.
• Assigns specialized roles (like researcher, coder, and editor) to different AI models.
• Passes the work back and forth between them until the final, polished result is ready.
• 分析任务并将其拆解为多个步骤。
• 为不同的 AI 模型分配专业角色（如研究员、程序员和编辑）。
• 在模型之间反复传递工作，直到最终产出完美的结果。

By working as a team, these models achieve far better results than any single AI could on its own. 通过团队协作，这些模型能够实现远超单一 AI 独立运作的效果。

The Secret Sauce: Teamwork Over Size

秘诀：团队协作胜过模型规模

Fugu’s intelligent coordination is based on two core concepts presented at the ICLR 2026 conference: Fugu 的智能协调基于 ICLR 2026 会议上提出的两个核心概念：

1. The Manager (TRINITY) 1. 管理者 (TRINITY)

Think of TRINITY as the manager of the team. It is a compact coordinator model that assigns specific roles to the worker LLMs. For example, it might tell Gemini to generate the initial code, tell Claude to find the bugs, and tell GPT to write the user documentation. They collaborate seamlessly without needing to merge their code bases. 可以将 TRINITY 视为团队经理。它是一个紧凑的协调模型，负责为工作 LLM 分配具体角色。例如，它可能会指示 Gemini 生成初始代码，让 Claude 查找 Bug，并让 GPT 编写用户文档。它们无需合并代码库即可实现无缝协作。

2. The Conductor 2. 指挥家 (The Conductor)

The Conductor acts as the communication network designer. It figures out the best way for the worker AIs to talk to each other for your specific question. It writes custom instructions for each worker model, discovering collaboration strategies that human designers might never have thought of. “指挥家”充当通信网络设计师的角色。它会针对你的具体问题，找出工作 AI 之间沟通的最佳方式。它为每个工作模型编写自定义指令，从而发现人类设计师可能从未想到的协作策略。

Why Fugu is the Future of AI

为什么说 Fugu 是 AI 的未来

Fugu represents a major shift in how we use artificial intelligence: Fugu 代表了我们使用人工智能方式的一次重大转变：

• It is Future-Proof: When a new, more powerful LLM is released, Sakana AI simply adds it to Fugu’s model pool and updates the coordinators. Fugu gets smarter automatically, while your API endpoint remains exactly the same.
• Superior Reasoning: Collaborative AI excels at complex, multi-step problem solving where single models often get lost or make mistakes.
• Easy to Use: You don’t need to know how to build complex agent networks. You just call one API, and Fugu handles all the heavy lifting.
• 面向未来： 当发布更强大的新 LLM 时，Sakana AI 只需将其添加到 Fugu 的模型池并更新协调器即可。Fugu 会自动变得更聪明，而你的 API 端点保持不变。
• 卓越的推理能力： 协作式 AI 擅长解决复杂的、多步骤的问题，而单一模型往往会在这些问题上迷失方向或出错。
• 易于使用： 你无需了解如何构建复杂的智能体网络。只需调用一个 API，Fugu 就能处理所有繁重的工作。

Getting Started with Sakana Fugu

如何开始使用 Sakana Fugu

For developers on macOS or Ubuntu, you can install Fugu locally with a single terminal command: 对于 macOS 或 Ubuntu 系统的开发者，可以通过终端运行以下命令在本地安装 Fugu：

curl -fsSL https://sakana.ai/fugu/install | bash

Once installed, you can start it by running: 安装完成后，运行以下命令即可启动：

codex-fugu

The Verdict

总结

Sakana Fugu is proof that the future of AI isn’t just about building bigger models—it’s about teaching AI how to work together. By organizing different models into a cooperative team, Fugu delivers a smarter, faster, and more accessible AI experience for everyone. Sakana Fugu 证明了 AI 的未来不仅仅在于构建更大的模型，更在于教会 AI 如何协同工作。通过将不同的模型组织成一个协作团队，Fugu 为每个人带来了更智能、更快速、更易用的 AI 体验。

Want to dive deeper? Check out the Sakana Fugu GitHub Repository and read the arXiv Technical Report. 想深入了解吗？请查看 Sakana Fugu 的 GitHub 仓库并阅读 arXiv 技术报告。