🌻Buddie
Science & TechMoneyJobsHealthPeopleProfileDonate
Post
Back to Money
AIJune 22, 2026 · 4 min read

🤖 Japan's Sakana AI Unveils Fugu, a Multi-Agent System That Orchestrates Rival Frontier Models

On June 22, 2026, Tokyo-based Sakana AI released Fugu, a multi-agent orchestration system that coordinates multiple frontier models behind a single API. Rather than building one bigger model, Sakana sells the layer that routes work across GPT-5, Gemini 3.1 Pro, and Claude Opus 4.8 — and claims the result rivals Anthropic's Fable and Mythos.

On June 22, 2026, Tokyo-based Sakana AI released Fugu, a multi-agent orchestration system that coordinates multiple frontier large language models behind a single, OpenAI-compatible API. Rather than competing to build the single most powerful model, Sakana AI is selling the layer that coordinates them — turning orchestration from a developer's engineering problem into a turnkey product.

At Fugu's core is a 7-billion-parameter model trained via reinforcement learning, dubbed the "RL Conductor." Notably, this small model does not generate the final answer itself. Instead, it acts as a foreman: when a user submits a task, the RL Conductor dynamically analyzes the task type and delegates sub-tasks to the most suitable specialists in its agent pool — top-tier models such as GPT-5, Gemini 3.1 Pro, and Claude Opus 4.8 — before assembling their contributions into a single coherent response.

Sakana AI reports that Fugu surpasses GPT-5.5 and Claude Opus 4.8 on SWE-Bench Pro and TerminalBench, two demanding benchmarks for software engineering and command-line agentic tasks, and claims the system rivals Anthropic's Fable and Mythos models. If borne out by independent evaluation, the results suggest that intelligent orchestration of existing models can rival, and in some cases exceed, the performance of any single frontier system.

The approach carries a strategic dimension. By coordinating rather than replacing the major labs' models, Fugu offers customers a hedge against vendor lock-in: workloads can be routed across providers, and the system can adapt as the relative strengths of individual models shift. For enterprises wary of betting on a single AI supplier, this orchestration layer is an appealing proposition.

Sakana AI was founded in Tokyo in 2023 by Llion Jones — a co-author of the seminal "Attention Is All You Need" Transformer paper — and former Google researcher David Ha. Now valued at over $2.5 billion, it is Japan's most valuable unlisted AI unicorn, and Fugu represents its most ambitious bid yet to carve out a distinctive position in a market dominated by U.S. labs racing to build ever-larger standalone models.

Published on Buddiespace · June 22, 2026

More from Money 💰

🐡

Fugu Faces Early Scrutiny: Users Report Slow Speeds and Heavy Token Use

June 24, 2026 · 3 min read

📉

Global Tech Sell-Off Deepens as AI and Chip Stocks Tumble

June 23, 2026 · 4 min read

🏠

China's Housing Market: From Boom to Reset

June 23, 2026 · 5 min read

🛸

SpaceX Acquires Cursor in $60 Billion Deal, Merging Space and AI

June 16, 2026 · 5 min read

🏥

Abridge Partners with NVIDIA, Expands Beyond Clinical Documentation

June 14, 2026 · 5 min read

🚀

SpaceX IPO: The Most Anticipated Listing in a Generation

June 13, 2026 · 5 min read

🤖

Anthropic Eyes IPO as AI Valuations Soar

June 9, 2026 · 3 min read

🇨🇳

DeepSeek Set to Raise $7.5 Billion as Chinese AI Rival Reshapes Global Markets

June 9, 2026 · 3 min read

📖

Ka-Shing Li is promoting his autobiography in Chinese social media

June 9, 2026 · 3 min read

💡

OpenAI IPO: What Investors Need to Know

June 9, 2026 · 4 min read