Asian AI Startups Pitch Frontier Models Around US Export Controls
TechCrunch reported that Sakana AI and 360 are promoting Mythos-like models as US restrictions on Anthropic systems create a market opening for Asian frontier AI providers.
Asian Model Launches Meet An Export-Control Opening
Asian AI startups are using US model-access restrictions as a commercial opening.
TechCrunch reported that Chinese cybersecurity firm 360 unveiled Tulongfeng, an AI tool it says can compete with Anthropic's Mythos.
Earlier in the same week, Tokyo-based Sakana AI launched Fugu, a frontier model named after the Japanese word for blowfish.
Sakana said the model stands shoulder-to-shoulder with leading systems such as Anthropic's Fable 5 and Mythos Preview.
Sakana Frames Fugu For Agents
Fugu is designed for agents and can orchestrate access to other models through their APIs.
Sakana says it is targeting Japanese businesses and government agencies that want frontier capability while reducing exposure to tightening export controls.
The company said the launch timing was coincidental, but its website advertised frontier capability without export-control risk.
A Sakana spokesperson said the product had been in development since last year and that the research behind it was presented at ICLR in the spring.
Sakana Says US Models Remain Important While Buyers Seek Optionality
Sakana did not claim that Asian customers are abandoning US AI providers.
The company said US models remain important to Asia, while describing the current moment as one in which buyers want resilience and optionality.
TechCrunch reported that David Ha, Llion Jones and Ren Ito launched Sakana in 2023.
Ha and Jones previously worked at Google, while Ito had roles at Mercari and Stability AI.
The company focuses on lower-cost generative AI systems tuned for Japanese language and cultural requirements and designed to work with smaller datasets.
Capability Claims Need Independent Benchmarks
The source places the launches against the US decision to restrict Anthropic's Mythos and Fable 5 models for non-Americans.
Regional providers are marketing model availability as a feature while those restrictions limit access for non-American users.
The public record does not yet include independent benchmark results for Tulongfeng or Fugu against Mythos, Fable 5 or other frontier systems.The public record still lacks customer names, pricing, deployment dates or government procurement commitments for the new models.
