Sakana Fugu is a multi-agent system that Sakana AI ships as one model behind a single API. When the user sends a prompt, Fugu plans the task, picks which models in its pool of frontier LLMs to call, runs them in parallel, verifies the partial results, and returns one combined answer.

How does Fugu Ultra compare to Fable 5?

Sakana AI says Fugu Ultra stands shoulder-to-shoulder with Anthropic's Fable 5 across coding, reasoning, science, and agentic benchmarks, including AutoResearch, Rubik's Cube solving, mechanical design, and financial time-series prediction. Sakana AI ran the comparison; independent benchmarks have not yet been published.

How is Fugu Ultra different from Fugu?

Fugu is the everyday model — tuned for low latency, meant for coding tools, chatbots, and interactive services. Fugu Ultra coordinates a deeper pool of expert agents and prioritizes answer quality on harder, longer, higher-stakes work like paper reproduction, cybersecurity analysis, and patent investigation.

Where can I use Sakana Fugu?

Sakana Fugu is generally available through the Sakana AI console at console.sakana.ai with an OpenAI-compatible API for Chat Completions and Responses. The service is not available in the EU/EEA at launch. Subscription tiers cover everyday use and a pay-as-you-go plan handles heavier or enterprise workloads.

Can I exclude specific models from Fugu's pool?

Yes. Fugu lets the caller exclude specific agents from the orchestration pool to meet compliance constraints, which matters for teams that cannot route requests through certain providers (e.g. Fable 5 under US export controls). The exclusion happens at the API layer rather than requiring a separate deployment.

Sakana AI · 2026-06-22 · major

Sakana Fugu — multi-agent orchestration model that matches Fable 5 on quality

Sakana AI launched Fugu and Fugu Ultra, a multi-agent orchestration model delivered as one OpenAI-compatible API. Fugu Ultra coordinates a pool of expert agents and is reported to match Fable 5 on coding, reasoning, science, and agentic benchmarks.

Sakana AI ships Fugu, a single API that routes each request to a pool of frontier models and verifies the answer before returning it.

Quick facts

Maker	Sakana AI (Tokyo)
Variants	Fugu, Fugu Ultra
Architecture	Multi-agent orchestration over a pool of LLMs
API	OpenAI-compatible (Chat Completions + Responses)
Availability	GA via console.sakana.ai; not in EU/EEA
Foundations	TRINITY + Conductor (ICLR 2026)
Pricing	Subscription tiers plus pay-as-you-go for heavier use

What is it?

Sakana Fugu is a multi-agent system delivered as one model. Behind a single OpenAI-compatible endpoint, Fugu plans a task, picks which models in its pool to call, splits work across them, verifies the partial results, and composes a final answer. Two variants ship at launch: Fugu for everyday low-latency work, and Fugu Ultra for high-stakes multi-step problems.

How does it work?

Two Sakana AI research lines feed Fugu. TRINITY (ICLR 2026) evolves an LLM coordinator that splits a task into Thinker, Worker, and Verifier roles. Conductor (ICLR 2026) trains a model with reinforcement learning to write natural-language coordination plans for a pool of LLMs. Fugu combines both: it acts as the orchestrator language model, calling other LLMs (and instances of itself) instead of merging weights.

Why does it matter?

Fugu turns 'pick a model and a framework' into one API call, which lowers integration cost for teams building agents. Sakana AI says Fugu Ultra stands shoulder-to-shoulder with Anthropic's Fable 5 — relevant because Fable 5 is currently banned from US export, so an orchestration model that can route around any one supplier is structurally useful for enterprises affected by the ban.

Who is it for?

Coding-agent and research-workflow builders who want one endpoint, not five

Frequently asked questions

What is Sakana Fugu?: Sakana Fugu is a multi-agent system that Sakana AI ships as one model behind a single API. When the user sends a prompt, Fugu plans the task, picks which models in its pool of frontier LLMs to call, runs them in parallel, verifies the partial results, and returns one combined answer.
How does Fugu Ultra compare to Fable 5?: Sakana AI says Fugu Ultra stands shoulder-to-shoulder with Anthropic's Fable 5 across coding, reasoning, science, and agentic benchmarks, including AutoResearch, Rubik's Cube solving, mechanical design, and financial time-series prediction. Sakana AI ran the comparison; independent benchmarks have not yet been published.
How is Fugu Ultra different from Fugu?: Fugu is the everyday model — tuned for low latency, meant for coding tools, chatbots, and interactive services. Fugu Ultra coordinates a deeper pool of expert agents and prioritizes answer quality on harder, longer, higher-stakes work like paper reproduction, cybersecurity analysis, and patent investigation.
Where can I use Sakana Fugu?: Sakana Fugu is generally available through the Sakana AI console at console.sakana.ai with an OpenAI-compatible API for Chat Completions and Responses. The service is not available in the EU/EEA at launch. Subscription tiers cover everyday use and a pay-as-you-go plan handles heavier or enterprise workloads.
Can I exclude specific models from Fugu's pool?: Yes. Fugu lets the caller exclude specific agents from the orchestration pool to meet compliance constraints, which matters for teams that cannot route requests through certain providers (e.g. Fable 5 under US export controls). The exclusion happens at the API layer rather than requiring a separate deployment.

Try it

https://console.sakana.ai/login