Sakana Fugu AI: The Multi-Agent API Explained

Q: Where do I get the API?

Sign up at sakana.ai - one key covers the whole model panel.

Sakana Fugu is a multi-agent AI API from Japanese lab Sakana — one API call runs a whole panel of models (closed and open), which get synthesised into a single answer, with a flagship tier, Fugu Ultra, that reportedly matches frontier models like Fable and Mythos.

Think Fusion, but faster and around a quarter of the cost. Here's the Sakana Fugu AI multi-agent API explained.

The One-Line Answer

Sakana Fugu = a multi-agent panel behind a single API. You call it once, a panel of models answers, and you get one synthesised result — no juggling individual model providers.

"Fugu" is Japanese for pufferfish; Sakana is the lab (founded by ex-Google researchers, including a co-author of "Attention Is All You Need").

How The API Works

Send a prompt to one endpoint. Sakana automatically manages model selection and delegation — it decides which models in the panel run, executes them, and fuses the answers.

That's the convenience: one API key handles everything. You don't sign up to each underlying model; Sakana orchestrates the panel for you, mixing closed and open models.

It's one-shot — you wait for the panel to respond, rather than going back and forth like a Claude CLI session.

This is the same pattern as OpenRouter Fusion, which is why I've built both into my Agent OS side by side.

Fugu vs Fugu Ultra

	Fugu	Fugu Ultra
Focus	Low latency, speed	Maximum answer quality
Best for	Coding (Codex), customer-facing	Hard multi-step tasks, AI research
Cost	Lower	Higher (flagship)

Fugu's speed is the big edge over Fusion, which can take 5–10 minutes for an answer. Fugu Ultra is the one that reportedly stands shoulder-to-shoulder with Fable and Mythos.

🔥 Want Sakana ready to use in an agent dashboard? The AI Profit Boardroom — the award-winning #1 AI community — has Sakana and Fusion wired into the Agent OS already, with daily updates and 5 coaching calls a week. → Get the Agent OS + Sakana System

Benchmarks (Read With Care)

On Sakana's published numbers, Fugu is even with or beating Fable 5 on most benchmarks — Terminal Bench (Fugu Ultra 82.1 vs Fable 5 80.4), Live Code Bench (Fugu Ultra 93.2) — except SW Bench Pro, where Fable 5 wins clearly.

But these are self-reported. We've watched labs game their own benchmarks before (a recent viral one fooled a lot of smart people). So verify independently.

That's what Goldie Bench is for — my own model leaderboard. Fugu Ultra scores 8.62/10 there (3 golds) — see the full Goldie Bench scorecard and judge the real outputs yourself.

Output Quality vs The Big Models

Running the same prompts I use on every model (website, maze game, galaxy sim, orbit sim), Fugu's outputs looked excellent — clean UI, strong animations.

Side by side it edged out GLM 5.2 (buggier), Opus 4.8 (duller UI), and even Fusion (marginal, but nicer design and motion).

Cost vs Fusion

Fusion (OpenRouter): pay-per-use, API only, expensive.
Sakana Fugu: ~25% of Fusion's cost for the same prompts, plus a flat-rate subscription option.

For high-volume agent loops, that flat rate is the deciding factor.

How To Use It

Sign up at sakana.ai for an API key.
Pick Fugu (fast) or Fugu Ultra (deep quality).
Call the API or drop it into a tool like Free Claude Code or the Agent OS.

Note: at launch, some users reported it wasn't available in the EU/UK — check your region.

Frequently Asked Questions

What is the Sakana Fugu API?

A single API that runs a multi-agent panel of models (closed and open), synthesising one answer. It manages model selection and delegation for you, with a premium tier called Fugu Ultra.

How is it different from Fusion?

Same panel-and-synthesis concept, but lower latency, roughly 25% of the cost, and a flat-rate subscription option.

Fugu or Fugu Ultra — which should I use?

Fugu for fast, lower-cost work (coding, customer-facing); Fugu Ultra for hard, multi-step, high-quality tasks.

Are the benchmarks trustworthy?

They're self-reported, so treat them cautiously. On the independent Goldie Bench, Fugu Ultra scores 8.62/10 across scored tasks.

Where do I get the API?

The Verdict

Sakana Fugu AI makes the multi-agent panel practical: one API, frontier-level outputs, faster and cheaper than Fusion, with a flat rate for heavy use. Verify the hype on your own tasks — but Sakana Fugu AI is a genuinely exciting drop.