Japan’s Sakana AI has launched Fugu, a multi-model orchestration system that matches Anthropic’s export-restricted Claude Fable 5 and Mythos on benchmarks—without using either. Fugu pools public models like Gemini 3.1 Pro, GPT-5.5, and Claude Opus 4.8, picking and combining them per task through one API. CEO David Ha pitches it as a hedge against single-vendor dependence after US export controls cut off Anthropic’s top models. Critics call it a closed orchestrator on closed models.