Sakana AI Introduces Fugu, a Model Orchestrator for Cost and Performance

Summary

Sakana AI presents Fugu as a model that orchestrates leading large language models. Its stated goal is to push the Pareto frontier between cost and performance by coordinating different models rather than relying on one system. Fugu is designed to keep operating if a closed model becomes unavailable, adding resilience to the orchestration approach. Sakana AI says the variety of tasks handled by the participating models makes Fugu particularly suited to agentic chatbots. The company highlights three versions: Fugu Ultra for maximum performance, Fugu Cyber for cybersecurity, and Fugu Max for the cost-performance Pareto frontier. Fugu can be tried through Sakana Chat.