Skip to the document
Madhuopen lab
Outcome School · AI agents31 min read

by Amit Shekhar · 22 June 2026

Decoding Sakana Fugu Technical Report

In this blog, we are going to learn about Sakana Fugu, a family of AI models that work like a conductor for a team of other AI models.

6,073 words#ai-agent#llm#ai#machine-learning4 recall cards

Decoding Sakana Fugu Technical Report
Read on Outcome School ↗then come back to lock it in
Before you read, guess

Why was Fugu necessary compared to older fixed teams and simple routers?

Ten seconds, a guess, then read — a wrong guess still makes the answer stick.

What this article covers

  1. What is Sakana Fugu?
  2. Why Fugu was needed
  3. The big picture: what Fugu does
  4. Collective Intelligence
  5. The two Fugus - Fugu and Fugu-Ultra
  6. How Fugu picks the right model - the lightweight selection head
  7. Teaching Fugu who is best - supervised fine-tuning
  8. Polishing Fugu on real tasks - evolutionary strategies
  9. How Fugu-Ultra conducts an orchestra - the Conductor
  10. Teaching Fugu-Ultra to conduct - GRPO
  11. Stopping the agents from copying each other
  12. How well does Fugu perform?
  13. The clever strategies Fugu discovered on its own

The article lives on outcomeschool.com. Read it there, then come back: the tutor in the margin has read it and will answer questions, and the questions below check what stayed.

Before you go

In one sentence, what was this chapter about?

From memory, without scrolling up. Writing it is what makes it yours; the grade is only to show you what you had.

How sure?