Sakana AI Fugu: Orchestrating LLMs, Matching Anthropic

Sakana AI is launching Fugu, a new system that coordinates multiple language models to act like a single model. This system matches the benchmarks of Anthropic's Fable and Mythos, even though those models are not part of Fugu's internal pool. Fugu comes in a base version for daily tasks and a more powerful Fugu Ultra variant. Fugu Ultra performs on par with top-tier models in coding, reasoning, and science benchmarks. The system's design also aims to reduce reliance on any single AI provider. What's interesting is that Fugu itself is a language model, trained to call other LLMs from an agent pool. This allows it to handle tasks directly or bring together a team of specialized models as needed. This innovation could offer more flexible and resilient AI solutions.