Perplexity AI Describes Its Embedding Serving Infrastructure

Perplexity AI posted from its official account on X that its serving infrastructure lowers latency and improves throughput for both online and batch embedding workloads. The company stated that combining its Ivy, Tulip, and ROSE systems produces faster search at reduced cost compared with off-the-shelf solutions. The post included two charts, the first labeled Embeddings Request Latencies, as supporting visuals for the performance claims.