Sakana AI's orchestrator adds Nvidia Nemotron to prove "collective intelligence" can rival single frontier models

Sakana AI has integrated Nvidia’s Nemotron-340B model into its Fugu orchestrator, a system designed to dynamically combine multiple specialized language models to complete complex tasks. By demonstrating that coordinated smaller models can compete with larger frontier systems, the company is testing the viability of collective intelligence architectures over single-model development. This approach suggests a shift in AI strategy, where developers may prioritize efficient model orchestration to achieve performance levels previously restricted to massive, monolithic proprietary systems.