Sakana AI launches Fugu, a multi-LLM orchestrator that matches Anthropic's top models

Tokyo based Sakana AI launched Fugu, a multi-LLM orchestration system, on June 22, 2026. The system dynamically coordinates models from a swappable pool while presenting a single API to users, and its benchmark scores match Anthropic's Fable 5 and Mythos Preview.

Fugu Ultra scored 73.7 on SWE Bench Pro, outperforming Opus 4.8 (69.2) and GPT 5.5 (58.6) according to Sakana's published benchmarks. The base Fugu model targets everyday tasks with low latency, while Fugu Ultra is built for complex, multi-step problems.

Two variants target different workloads

Sakana AI is offering two versions of Fugu. The base model handles coding, code review, and chatbot use cases with low latency. Teams with privacy or compliance needs can exclude specific agents from the pool. Fugu Ultra is designed for maximum answer quality on complex tasks such as AI research, reproducing scientific papers, cybersecurity analysis, and patent searches.

• About 500 beta users tested the system in real world settings, according to Sakana AI.
• One software developer reported that Fugu Ultra caught more than 20 bugs during code review, compared to about three flagged by GPT 5.5.
• Fugu Ultra is now available through Vercel's AI Gateway, as announced on the Vercel blog.
• Both variants are accessible via a single OpenAI compatible API on Sakana's product page and console.
• Sakana offers subscription plans for daily use.

Orchestration as a hedge against vendor lock in

Sakana AI is pitching Fugu as a safeguard against single provider dependence. The company points to recent export controls on Anthropic's Fable and Mythos models as a concrete example of how access to top AI systems can vanish overnight due to regulatory shifts or foreign policy decisions. "For an organization or a nation, relying on a single company's APIs for critical infrastructure, finance, or governance is a material vulnerability. This risk is no longer a hypothetical possibility, but a reality," Sakana AI wrote in its announcement.

Fugu's model pool is fully swappable, so the system can reroute to other models if one provider goes dark. However, real world performance depends entirely on which models are in the pool. If several top providers restrict access simultaneously, Fugu's options shrink too. An orchestrator like Fugu may boost resilience, but it is not the same as true sovereignty. Still, early testers report gains on complex workflows. Sakana AI also claims Fugu beat Gemini 3.1 Pro, Opus 4.8, and GPT 5.5 in its own tests on automated research, mechanical design, and financial forecasting. Both variants are live now, and Sakana AI says the system is worth watching on performance alone.