The archive · AI & Models · Product decision · 2025–2026
Sakana AI launches Fugu, an orchestration model hedging Japan against US export controls
Tokyo startup Sakana AI ships Fugu, a model orchestrating other models' APIs, betting export controls make single-provider AI dependence untenable for Japan
Sakana AI
What the business is
Sakana AI builds affordable generative AI models for Japan, optimized for small datasets and Japanese language and culture; Fugu, launched late June 2026, is a single frontier model that orchestrates a pool of other models' APIs for agents.
How it started
Sakana AI was co-founded in Tokyo in 2023 by Google alumni David Ha and Llion Jones plus Ren Ito, formerly of Mercari and Stability AI, to make affordable generative AI models that work well with small datasets and are optimized for Japanese language and culture. Fugu was built from this conviction that the most powerful AI is a collaborative ecosystem, not an isolated monolith; Sakana said it had been building the model since 2025 and presented the research at ICLR in spring 2026.
What happened
Sakana AI closed a $135M Series B at a $2.65B valuation in November 2025 (about $379M raised in total per PitchBook via TechCrunch). In June 2026 the US government ordered Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national — access Sakana said can 'disappear overnight.' It launched Fugu in late June 2026 as two models, Fugu and Fugu Ultra, behind one OpenAI-compatible API, and described Fugu as more than a land grab: a model trained to delegate to, verify and combine other models, with a swappable agent pool that routes around any provider restriction. Sakana said it is targeting Japanese businesses and government agencies seeking to reduce exposure to tightening export controls, while insisting US models remain important to Asia and the moment is not a permanent realignment.
No ending yet — it is still running.
Background
Sakana AI is a Tokyo startup founded in 2023 by Google alumni David Ha and Llion Jones and former Mercari/Stability AI executive Ren Ito to build affordable generative AI models that work well with small datasets and are optimized for Japanese language and culture. In late June 2026 it launched Fugu, a frontier model it says stands shoulder-to-shoulder with Anthropic's Fable 5 and Mythos Preview, named after the Japanese word for blowfish.
The launch was a bet on orchestration: Fugu is itself a language model trained to decide when to delegate, and it coordinates a pool of other models' APIs — with a swappable agent pool — so customers get one model endpoint and multi-agent behavior underneath. Sakana argues that 'orchestration models are the next frontier, beyond bigger models' and that relying on a single AI provider for national infrastructure is a risk the recent export controls made impossible to ignore.
The trigger was political: in June 2026 the US government ordered Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national, cutting non-US customers off from two of the most capable frontier models. Sakana launched Fugu the same week Chinese cybersecurity firm 360 unveiled rival tools, and its marketing led with 'frontier capability without the risk of export controls.' Co-founder Ren Ito separately urged the US to preserve access for its closest allies, arguing AI should not be hoarded.
Sakana positioned Fugu as a hedge that preserves access to frontier AI rather than a replacement for US models — it said US models remain important to Asia — while targeting Japanese businesses and government agencies seeking to reduce exposure to tightening export controls. The company had closed a $135M Series B at a $2.65B valuation in November 2025, about a year after its Series A.
What has to be true
- Export controls turned a theoretical single-vendor risk into a lived event for Asian enterprises and governments, creating demand Sakana did not have to manufacture.
- Orchestration fit Sakana's founding thesis — collective ecosystems over isolated monoliths — so Fugu extended the company's research direction instead of chasing Anthropic's architecture.
- Positioning as a hedge rather than a replacement matched real customer sentiment, since Asian customers still wanted US models and would distrust a premature 'US AI is finished' pitch.
- The swappable agent pool made the product the argument: if one provider's access disappears, Fugu routes around it, turning export control from an existential risk into a configuration change.
What can be applied
When regulators cut a dominant supplier out of a market, local rivals win by selling the same capability plus local fit — and by framing themselves as a hedge, not a replacement.
Aftermath
As of 2026-06-27, Fugu was generally available through one OpenAI-compatible API in two tiers — Fugu for everyday use and Fugu Ultra for long multi-step work — and Sakana was aiming it at Japanese businesses and government agencies trying to reduce exposure to tightening export controls. The company said the agent pool is swappable by design and planned to fold in newer models including open models and its own. Sakana continued to describe the moment as a hedge: US models remain important to Asia, and the current realignment is not proclaimed as permanent.
Sources
- Asian AI startups launch Mythos-like models as Anthropic's export ban drags on
- Asian AI startups launch Mythos-like models as Anthropic's export ban drags on
- Sakana AI raises $135M Series B at a $2.65B valuation to continue building AI models for Japan
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card