Sakana AI's Fugu orchestrates multiple LLMs to match Anthropic's Fable and Mythos benchmarks

A Japanese startup has demonstrated that orchestrating multiple smaller language models can match the performance of Anthropic's cutting-edge systems, challenging the assumption that bigger models are inherently better. Sakana AI's approach, called Fugu, suggests the future of artificial intelligence may depend less on scaling up individual systems and more on intelligently coordinating them—a potentially more efficient path for companies that lack the resources of tech giants. The findings hint at a fundamental shift in how the industry might build competitive AI, with implications for both startups seeking to compete and established players defending their advantages.

The Decoder·2h ago