Japan’s Sakana Fugu Beats Opus 4.8 and GPT-5.5 by Conducting Them, Not Replacing Them

2026-06-26 · Source: Towards AI - Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Emerging Technologies & Innovation · Depth: Intermediate, medium

Summary

Sakana AI, a Tokyo lab, introduced Fugu on June 22, 2026, a novel system that orchestrates leading frontier models like OpenAI's GPT-5.5, Google's Gemini 3.1 Pro, and Anthropic's Claude Opus 4.8, rather than replacing them. Fugu is a small model designed to route tasks, delegate work to the best-suited specialist model, and synthesize their outputs. Benchmarks show Fugu-Ultra, its high-quality variant, surpasses individual models on most demanding tests. For instance, on SWE-Bench Pro, Fugu-Ultra scored 73.7, outperforming Opus 4.8 (69.2), GPT-5.5 (58.6), and Gemini 3.1 Pro (54.2). Similarly, on Terminal Bench 2.1, it achieved 82.1, exceeding GPT-5.5 (78.2), Opus 4.8 (74.6), and Gemini (70.3). While these are vendor-reported figures and not a complete sweep—Fable 5 still leads SWE-Bench Pro—Fugu demonstrates that coordinating existing models can achieve frontier-level performance.

Key takeaway

For AI Engineers or Directors of AI/ML building advanced systems, this demonstrates that your focus shouldn't solely be on acquiring the largest single models. Instead, you should explore orchestration frameworks like Sakana AI's Fugu, which coordinate existing frontier models to achieve superior performance on complex tasks. Evaluate such multi-model approaches against your specific workloads, as they offer a validated path to frontier-level results, potentially optimizing both performance and cost by leveraging existing powerful APIs.

Key insights

Orchestrating multiple frontier models can yield performance exceeding individual top-tier models.

Principles

• Orchestration can outperform single large models.
• System-level intelligence is distinct from model scale.
• Dynamic coordination enhances overall AI capability.

Method

A small model plans, routes sub-tasks to specialist frontier models, then synthesizes their responses into a single coherent answer.

In practice

• Implement multi-model architectures for complex problems.
• Consider dynamic model activation for cost efficiency.
• Employ cross-checking for high-quality synthesis.

Topics

• AI Orchestration
• Multi-model Systems
• Sakana AI Fugu
• Frontier Models
• Performance Benchmarking
• Agentic AI

Best for: CTO, VP of Engineering/Data, AI Architect, AI Scientist, AI Engineer, Director of AI/ML

Related on AIssential

• No Claude Fable 5? No problem: Sakana achieves frontier performance with new Fugu multi-model, auto synthesis system — Multi-agent orchestration provides resilience and competitive performance by dynamically routing tasks to specialized AI agents.
• Sakana’s Fugu takes aim at the frontier — Sakana AI's Fugu uses multi-agent orchestration to deliver frontier AI capabilities and hedge against geopolitical export controls.
• Japan’s Sakana AI Says Its New Tool Rivals Anthropic’s Fable 5 — Sakana AI's Fugu unifies diverse AI models via one interface, simplifying complex AI task execution.
• Sakana AI Wrapped an Entire Multi-Agent System Into One API (And It Beats Frontier Models on… — Sakana AI's Fugu simplifies multi-agent orchestration into a single API, achieving frontier model performance on benchmarks.
• The Future of AI Isn’t a Genius. It’s a Manager. — Intelligence can emerge from coordinating specialist AIs, not solely from a single, larger model.
• Sakana AI's Fugu orchestrates multiple LLMs to match Anthropic's Fable and Mythos benchmarks — Fugu orchestrates multiple LLMs to match top benchmarks and reduce single-provider dependence.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Towards AI - Medium.