Sakana AI orchestrates LLMs with RL Conductor

Sakana AI has developed the RL Conductor, a novel reinforcement learning-trained model designed to dynamically orchestrate multiple large language models (LLMs). This system overcomes the limitations of rigid, human-coded pipelines by analyzing inputs and distributing tasks among specialized LLMs, achieving superior performance on complex reasoning and coding benchmarks. The Conductor outperforms individual frontier models like GPT-5 and Claude Sonnet 4, as well as existing multi-agent frameworks, at a significantly lower cost and with fewer API calls. Sakana AI is commercializing this technology through its Fugu product, offering an API-based multi-agent orchestration service for enterprises.