What Is Sakana Fugu AI? (Fugu vs Fugu Ultra)

Q: How do I access Sakana Fugu?

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

Sakana Fugu is a multi-agent AI system from Japanese lab Sakana — you call one API, a panel of models (closed and open) answer behind the scenes, and you get a single synthesised result. Its flagship tier, Fugu Ultra, reportedly matches frontier models like Fable and Mythos.

Here's what Sakana Fugu AI is, the difference between Fugu and Fugu Ultra, and whether the hype holds up.

What Is Sakana Fugu?

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

Sakana Fugu is their multi-agent orchestration system, accessed through a single model API. Rather than one model answering, a panel of models — both closed and open — respond and get synthesised into one answer.

It's the same concept as OpenRouter Fusion: a panel plus a synthesis step. The differences are speed, price, and that it comes in two tiers.

How It Works

You send a prompt to one endpoint. Sakana handles model selection and delegation automatically — it picks the panel, runs it, and fuses the result. One API key; no signing up to each underlying model.

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

Fugu vs Fugu Ultra

	Fugu	Fugu Ultra
Strength	Low latency, fast	Maximum answer quality
Best for	Coding (Codex), customer-facing	Hard multi-step tasks, AI research
Cost	Lower	Higher (flagship)
Matches	Strong	Reportedly Fable / Mythos

Fugu's speed is the headline edge over Fusion (which can take 5–10 minutes per answer). Fugu Ultra is the quality flagship.

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

Benchmarks — And A Health Warning

On Sakana's published numbers, Fugu is even with or ahead of 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 clearly wins.

But these are self-reported. Labs have gamed their own benchmarks before — a recent viral one fooled a lot of smart people. So verify independently.

That's why I built Goldie Bench, an independent leaderboard. Fugu Ultra scores 8.62/10 there (3 golds) — see the full scorecard and compare the real outputs against other models yourself.

Output Quality

Running the same prompts on every model (website, maze game, galaxy sim, orbit sim), Fugu's outputs looked excellent — clean UI, strong animations — edging out GLM 5.2, Opus 4.8, and even Fusion in side-by-sides.

Cost

Fusion (OpenRouter): pay-per-use, expensive.
Sakana Fugu: ~25% of Fusion's cost for the same prompts, plus a flat-rate subscription — ideal for high-volume agent loops.

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 plug it into 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 Sakana Fugu AI?

A multi-agent AI system from Japanese lab Sakana, accessed via one API. A panel of models answers and gets fused into a single result, with a flagship tier (Fugu Ultra) that reportedly matches frontier models.

What's the difference between Fugu and Fugu Ultra?

Fugu is low-latency and fast (great for coding and customer-facing tools); Fugu Ultra is the flagship, tuned for maximum quality on hard, multi-step tasks.

How is it different from Fusion?

Same panel-and-synthesis idea, but faster, around 25% of the cost, and with a flat-rate subscription option.

Are the benchmarks trustworthy?

They're self-reported, so verify them. On the independent Goldie Bench, Fugu Ultra scores 8.62/10.

How do I access Sakana Fugu?

The Verdict

Sakana Fugu AI makes the multi-agent panel fast and affordable: frontier-class outputs, ~25% of Fusion's cost, two tiers for speed or depth. Verify the benchmarks on your own tasks — but Sakana Fugu AI is one of 2026's most interesting releases.