Chinese cybersecurity firm 360 recently launched Tulongfeng, an AI tool it claims can rival Anthropic’s Mythos, while Tokyo-based Sakana AI introduced Fugu, a model designed to work with agents and APIs. These moves come as the U.S. export ban on Mythos and its restricted version, Fable 5, continues to limit global access. You’re watching a shift in the AI model competition, with Asian startups positioning themselves as viable alternatives to U.S.-based models.

Sakana AI’s Fugu and 360’s Tulongfeng are not just technical responses, they’re strategic plays in a growing race for AI dominance. This article will show you how these developments are reshaping the global AI landscape and what it means for businesses navigating export restrictions and model choices.

U.S. Export Ban Creates a Gap in Global AI Access

The U.S. export ban on Anthropic’s Mythos and Fable 5 has left a clear opening for Asian startups to fill the gap in global AI access. With these models now restricted to American users, companies like 360 and Sakana AI have moved quickly to offer alternatives that can perform similar tasks. This isn’t just about filling a void, it’s about building a competitive edge in the AI model competition.

Sakana AI’s Fugu, for example, is marketed as a direct alternative that avoids the risk of export controls. The company’s spokesperson acknowledged the timing was “entirely coincidental,” but the message is clear: there’s a demand for models that aren’t bound by U.S. restrictions. This shift is not just a response to the ban, it’s a calculated move to reshape the AI landscape.

Chinese and Japanese Startups Launch Mythos-Like AI Models

360’s Tulongfeng AI Model

Chinese cybersecurity firm 360 has launched Tulongfeng, an AI tool it claims can match the capabilities of Anthropic’s Mythos. This move positions 360 as a direct competitor in the AI model competition, especially as U.S. export controls limit access to Mythos and its variants. Tulongfeng is designed for cybersecurity applications, a sector where AI is increasingly used for threat detection and response. Unlike U.S. models, Tulongfeng is not subject to the same export restrictions, giving it an advantage in markets looking for alternatives.

Sakana AI’s Fugu Model

Sakana AI, a Tokyo-based startup, has introduced Fugu, a model named after the Japanese word for blowfish. Fugu is designed to work with agents and APIs, allowing it to integrate with other models and tools. The company says it can stand up to leading models like Anthropic’s Fable 5 and Mythos Preview. Sakana AI’s spokesperson emphasized that the release was not a direct response to the export ban but acknowledged the timing gave the product more visibility. Fugu is tailored for Japanese businesses and government agencies, offering a localized solution that avoids the risk of export controls.

What These Models Offer Businesses and Governments

Localization and Cultural Optimization

Sakana AI’s Fugu is specifically optimized for the Japanese language and culture, making it a better fit for local businesses and government agencies. This level of localization is a key differentiator, especially for organizations that need AI tools to understand and operate within specific regional contexts. Unlike generic models, Fugu is built to handle nuances in language, business practices, and user behavior that are unique to Japan. This makes it more effective for applications like customer service, regulatory compliance, and internal operations.

360’s Tulongfeng, while focused on cybersecurity, also benefits from being developed in China, where it can be fine-tuned to address local threats and regulatory environments. This kind of cultural and regional specificity is increasingly valuable for businesses that want AI tools that work well within their own markets, rather than requiring costly and time-consuming customization.

Compliance with Export Controls

One of the most immediate benefits of models like Fugu and Tulongfeng is their ability to avoid the restrictions imposed by the U.S. export ban. Sakana AI explicitly markets Fugu as an alternative that “avoids the risk of export controls,” which is a major concern for companies and governments looking to reduce their reliance on U.S.-based AI. This compliance advantage allows Asian organizations to deploy AI solutions without the legal and logistical hurdles that come with using models like Mythos and Fable 5.

For businesses that need to operate in regions with strict data laws or export restrictions, these models offer a clear path forward. They enable companies to maintain control over their data, avoid geopolitical risks, and ensure that their AI systems are aligned with local regulations. This is not just a technical benefit, it’s a strategic one that can impact long-term operations and risk management.

How These Models Compare to Anthropic’s Mythos

Performance and Capabilities

Sakana AI’s Fugu and 360’s Tulongfeng are designed to match the performance of Anthropic’s Mythos in specific areas, but they differ in focus and use cases. Fugu is optimized for cybersecurity and agent-based workflows, allowing it to integrate with other models via APIs. This makes it a strong contender for organizations that need AI to handle complex, multi-model tasks. Tulongfeng, on the other hand, is tailored for cybersecurity applications, offering threat detection and response capabilities that align with the needs of Chinese firms and government agencies.

Both models lack the broad general-purpose capabilities of Mythos, which is known for its advanced reasoning and language understanding. While they can perform well in niche domains, they are not yet positioned as all-in-one solutions for enterprise AI needs.

Accessibility and Cost

A key advantage of Fugu and Tulongfeng is their accessibility outside the U.S. export controls. Sakana AI’s spokesperson noted that Fugu is marketed as a “frontier capability without the risk of export controls,” which is a direct appeal to businesses and governments in Asia. This makes it an attractive option for organizations looking to avoid the restrictions placed on U.S. models.

In terms of cost, Sakana AI positions Fugu as an affordable alternative to more expensive U.S. models. However, concrete pricing details are not widely available, and the true cost-benefit for enterprise users remains to be seen. For now, the models offer a compelling mix of localization, compliance, and strategic positioning in the AI model competition.

The Future of AI Model Competition in Asia

Potential Shift in AI Leadership

With Asian startups like Sakana AI and 360 rapidly developing models that rival U.S. offerings, the balance of AI leadership is beginning to shift. These companies are not just filling a gap, they’re challenging the dominance of Western models in critical areas like cybersecurity and agent-based workflows. The export ban on Anthropic’s Mythos has accelerated this shift, creating an opportunity for Asian firms to establish themselves as credible alternatives on the global stage.

Sakana AI’s Fugu, for instance, is specifically designed for Japanese businesses and government agencies, showing how regional needs are shaping the next wave of AI innovation. This localized approach could give Asian models an edge in markets where cultural and linguistic nuances matter. As these models continue to evolve, they may no longer be seen as second-tier alternatives but as key players in the AI model competition.

Strategic Implications for Businesses

For businesses, the rise of Asian AI models means more choices, but also more complexity. Companies must now consider not only the performance of these models but also the geopolitical and regulatory risks involved. Sakana AI’s spokesperson noted that Fugu was built to avoid export control issues, which is a direct response to the Anthropic export ban. This suggests that businesses looking to avoid U.S. restrictions may increasingly turn to Asian models.

However, reliance on non-U.S. models may also introduce new dependencies and risks. While these models offer localized benefits, they may not yet match the scale, integration, or ecosystem support of established U.S. models. Businesses must weigh these factors carefully as they navigate the growing AI model competition in Asia.

Source: techcrunch.com