Perplexity launches Photon: Rust retrieval engine cuts p99 latency from 800 ms to 65 ms

Perplexity released Photon, an in-house retrieval and ranking engine written in Rust that replaces its forked open-source engine and now serves all production traffic. A new Fast Search mode in the Search API costs $1 per 1,000 requests with 160 ms p50 and 230 ms p95 latency.

• p99 retrieval and ranking latency dropped from about 800 ms to about 65 ms
• Fast Search reports 160 ms p50 and 230 ms p95 at $1 per 1,000 requests
• Photon runs on ~20% fewer serving machines and stores ~2.5x more data per document
• Fast Search cut estimated agent task cost by ~68% at comparable quality