Tokyo startup Sakana AI launches Fugu, an orchestrator that coordinates multiple AI models to match top competitors' performance while reducing reliance on any single vendor.

What happened: Sakana AI unveiled Fugu, a system that dynamically selects and coordinates multiple language models from a swappable agent pool to handle tasks. Two variants are now live: a base Fugu model for everyday performance, and Fugu Ultra for complex multi-step problems. According to Sakana's published benchmarks, Fugu Ultra performs on par with Anthropic's Fable 5 and Mythos Preview across coding, reasoning, science, and agent benchmarks.

Why it matters: Sakana is positioning Fugu as a safeguard against single-provider dependence. The company points to recent export controls on Anthropic's models as a concrete example of how access to top AI systems can vanish overnight due to regulatory shifts. Because Fugu's model pool is fully swappable, the system can reroute to other models if one provider becomes unavailable—though the real-world performance depends entirely on which models remain accessible.

What to watch: About 500 beta users have already tested Fugu in real-world settings, with early reports showing strength on long, multi-step workflows like automated data research, security analysis, and code reviews. Both variants are live now through a single OpenAI-compatible API; Sakana offers subscription plans for daily use and usage-based billing for larger workloads. One open question the announcement does not address: how much the orchestration drives up token usage and costs.