A Japanese artificial intelligence company, Sakana AI, says its new model, called Fugu, has matched the performance of Anthropic's Claude Fable 5. According to The Times of India, Sakana paired that claim with a public message aimed at countries that have been banned from using Anthropic's most powerful AI models.
What makes Fugu different, according to Financial Express, is its design. Rather than a single large "monolithic" system of the kind built by Western labs, Fugu is described as a multi-agent AI system — meaning it coordinates several smaller AI components to work together. Financial Express reports that this approach is meant to deliver agile, frontier-level performance while offering stronger privacy and what it calls geopolitical resilience.
That framing points to a bigger ambition. Financial Express says Fugu is positioned to help build "sovereign AI" — capabilities a country can run and control on its own, without depending on foreign providers. For nations shut out of the most advanced Western models, a homegrown or non-Western alternative that performs at a comparable level could matter a great deal.
The claim is drawing wider attention in the AI press. A report from 36Kr openly asks whether an AI model that surpasses Claude has now been born, a sign that Sakana's announcement is being taken seriously rather than dismissed.
It is worth noting that these are claims relayed through news coverage, and the sources here do not include independent benchmark data verifying that Fugu equals Claude Fable 5 across tasks. Readers should treat performance parity as asserted, not yet independently confirmed.
Why it matters: if a Japanese challenger can credibly match a leading US model while emphasizing privacy and national control, it signals that frontier AI is becoming a contest between competing visions — open versus restricted, monolithic versus multi-agent, and globally shared versus nationally sovereign.