Two weeks after the Trump administration effectively banned Anthropic's most powerful models from non-US users, Asian AI startups are stepping into the void with ambitious alternatives.
On June 22, Tokyo-based Sakana AI launched Fugu Ultra, a frontier-class model that matches Anthropic's Fable 5 and Mythos Preview on benchmarks while explicitly advertising "no risk of export controls." Three days later, Chinese cybersecurity giant 360 unveiled Tulongfeng, an AI tool the company claims can go head-to-head with Anthropic's Mythos for vulnerability discovery.
For developers and enterprises in Asia — and anyone outside the US who lost access to Anthropic's best models — this is the most significant shift in model availability since the export controls started.
What Actually Happened
On June 12, the US government ordered Anthropic to suspend all access to Claude Fable 5 and Claude Mythos 5 for anyone who isn't a US national — including Anthropic's own non-US employees. The move was unprecedented: a direct government intervention blocking access to a publicly released frontier AI model.
The impact was immediate. Companies and government agencies worldwide that had built workflows around these models lost access. On June 27, Mythos 5 was partially restored for a small group of approved US organizations, but Fable 5 remains unavailable, and the export restrictions themselves haven't been lifted.
This regulatory vacuum created a market opening — and Asian AI startups moved fast.
Sakana Fugu: Orchestration as a Hedge
Sakana AI's response isn't just another big model — it's a fundamentally different approach. Fugu is what the company calls an "orchestration model": a language model trained to call other models via API, routing tasks to the best specialist for each job. Think of it as an intelligent dispatcher that coordinates a team of expert models behind a single API endpoint.
Fugu Ultra, the flagship, matches or exceeds frontier benchmarks across coding, reasoning, and scientific tasks. The company's benchmark results show it competitive with Anthropic Fable 5 and Mythos Preview on engineering and reasoning benchmarks, while also handling agentic workflows through model orchestration.
Key technical differentiators:
• Single API, multi-model orchestration — You send one request; Fugu decides whether to answer directly or delegate to specialist models
• Recursive self-calling — Fugu can invoke itself as a sub-agent, enabling nested reasoning
• Vendor independence — If any model provider restricts access, Fugu dynamically routes around it
• OpenAI-compatible API — Drop-in replacement for existing OpenAI client code
• Privacy controls — Teams can opt specific agents out of the pool for compliance
The timing is deliberate but coincidental, according to Sakana. A spokesperson told TechCrunch the release was "entirely coincidental" with the export ban, though the company's website prominently advertises "delivering frontier capability without the risk of export controls."
Sakana co-founder David Ha framed it as a structural shift: "Orchestration Models are the next frontier, beyond bigger models." He argues that relying on a single provider for national infrastructure is a risk the export controls made impossible to ignore. "Collective intelligence is the practical hedge against this concentration of power."
360's Tulongfeng: Cyber Defense as National Strategy
Chinese security firm 360 took a different approach. Its Tulongfeng tool (named after a mythological Chinese dragon) is specifically designed for automated vulnerability discovery — the same cybersecurity sweet spot that made Anthropic's Mythos 5 so valuable.
According to Reuters, 360 also unveiled Yitianzhen, an AI system for automated cyber defense and incident response.
360's founder Zhou Hongyi didn't mince words. He described vulnerability-finding AI as a "national strategic asset" and warned of "one-way transparency" — a scenario where some nations have advanced detection capabilities while others don't. This is effectively a mirror of the same concern the US government used to justify its export controls.
What This Means for Developers
If you're building outside the US — or building for customers outside the US — the landscape just got more interesting:
More options, not fewer. Two weeks ago, if you wanted frontier-level AI for security or complex reasoning, your choices were effectively OpenAI, Anthropic, or Google. Now Sakana Fugu is a viable alternative, especially for teams that value vendor diversity.
Orchestration might be the smarter bet. Fugu's approach of coordinating multiple models raises a practical question: is a single monolithic model better, or is a system that dynamically picks the best tool for each subtask? For long-running agent workflows, orchestration may win.
Pricing is unannounced. Neither Sakana nor 360 has published detailed API pricing yet. Sakana offers subscription tiers and a pay-as-you-go plan for enterprise use. Developers should expect competitive pricing aimed at undercutting US models.
Export controls create market gravity. The US ban hasn't just inconvenienced non-US users — it's actively accelerated the development of local alternatives. Even if the ban gets lifted, Asian enterprises now have homegrown options trained on their languages and data. As Sakana's co-founder Ren Ito wrote: "AI sovereignty is about options, not ownership."
The Bigger Picture
Anthropic's run-rate revenue crossed $47 billion in May 2026. How much of that comes from Asian enterprise customers isn't public, but the numbers are large enough that losing even a fraction of that market hurts.
More importantly, this shifts the competitive dynamic. US AI labs that spent years building global distribution may find Asian markets increasingly difficult to access or retain. Meanwhile, Asian startups — backed by sovereign AI initiatives and rising domestic demand — have a clear path to growth.
The two-week period since the export ban may end up being one of those inflection points that looks obvious in retrospect. Asian alternatives were coming anyway. The ban just made them arrive faster, with a built-in sales pitch: "Can't be banned."
Sources
• TechCrunch: Asian AI startups launch Mythos-like models as Anthropic's export ban drags on
• Sakana AI: Fugu Release — One Model to Command Them All
• Reuters: China's 360 says it has developed tools to match Anthropic's Mythos
• The Verge: Anthropic's Mythos 5 is back
• Anthropic: Statement on US government directive
• Korea Times: AI sovereignty is about options, not ownership