# Sakana AI 发布 Fugu Ultra 多模型编排系统

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-06-22 12:12
- AIHOT 分数：70
- AIHOT 链接：https://aihot.virxact.com/items/cmqoq5ls903bkslx6xf3g4xfe
- 原文链接：https://x.com/rohanpaul_ai/status/2068909919873757456

## AI 摘要

Sakana AI 推出 Fugu Ultra，一个多智能体编排层，通过单一 OpenAI 兼容端点动态调用多个模型处理子任务。Fugu 本身是一个 LLM，被训练来自主决定是直接回答还是将子任务分发给模型池中的其他模型（包括递归调用自身），最后整合输出。在大多数基准测试中，Fugu Ultra 性能匹配 Fable 和 Mythos，提供前沿能力且规避出口管制风险。

## 正文

Sakana AI has unveiled Fugu Ultra， an orchestration layer that assembles and routes subtasks across a pool of models through one OpenAI-compatible endpoint.

On most benchmarks， it matches the performance of Fable and Mythos.

Fugu is a learned coordinator model inside a multi-agent system.

"Sakana Fugu is itself an LLM， trained to call various LLMs in an agent pool， including instances of itself recursively. Fugu dynamically orchestrates the world's best models to tackle complex， multi-step tasks."

When you send a prompt， Fugu decides whether to answer alone or hand pieces of the job to other models， then it gathers the outputs and produces one final response.

### 引用推文

> Sakana AI：Introducing Sakana Fugu: A full multi-agent orchestration system accessible via a single model API. Our 'Fugu Ultra' model matches the performance of Fable and ...