Case Study: Yutori achieves faster, lower-cost browser agent inference with Together AI

Yutori runs browser-use agents 2x faster and 4-5x cheaper with Together AI

Yutori, an AI-native startup building browser-automation agents, faced significant challenges with their inference infrastructure. Their proprietary Navigator model required an inference platform capable of handling long, continuously growing contexts, two distinct workloads (throughput-heavy background agents and latency-sensitive real-time interactions), and unpredictable traffic spikes. Standard cloud platforms were not built for this unique agentic loop, risking browser timeouts and limiting growth. They turned to Together AI and its Serverless Inference product to find a solution.

Together AI implemented a purpose-built inference platform optimized for Yutori's specific needs. The solution leveraged efficient caching for long inputs and was tuned for both high throughput and low latency. As a result, Yutori achieved 2x faster inference and a 4-5x reduction in cost compared to alternatives, alongside 99.9% uptime. Together AI's elastic scaling allowed Yutori to handle traffic surges seamlessly, providing a reliable foundation for their consumer and developer products.

Yutori