Sakhanda Wire
NVDA $153.73 -1.54% MSFT $506.30 -2.26% GOOGL $280.09 -1.28% META $619.28 -1.14% AMZN $256.62 -0.44%
← Back to the news

Perplexity launches Photon retrieval engine with lower production latency

Perplexity has introduced Photon, an in-house retrieval and ranking engine written in Rust. It replaces an open-source engine previously forked for the company’s AI-native search stack and now processes all production traffic. Photon also supports Fast Search in the Perplexity Search API, where clients can set search_type: "fast" on POST /search. The hosted option costs $1 per 1,000 requests; Photon itself is not open source and cannot be self-hosted.

The previous system’s production p99 latency was about 800 ms, while cold reads, index merges and recovery operations created additional delays. Photon distributes requests across shards, performs retrieval and multiple ranking stages, and uses adaptive posting lists, budgeted traversal, compact document records and batched asynchronous disk reads. Separate index-building and serving systems allow version changes without the earlier latency spikes.

Perplexity reports that Photon reduced retrieval and ranking p99 latency to about 65 ms and uses about 20% fewer serving machines. Fast Search records 160 ms at p50 and 230 ms at p95. Across 3,554 benchmark tasks, it achieved 64.3% at an estimated $59.73 model-plus-search cost, compared with 64.0% and $187.60 for the default preset. The faster mode had lower relevance and answer availability on internal long-tail tests.

Written by Sakhanda Wire AI from reporting by MarkTechPost. Facts, quotes and full context are in the original article below.

Originally published by MarkTechPost on

Read the original on MarkTechPost ↗

The reporting and the image belong to MarkTechPost and are credited and linked above; the text on this page is Sakhanda Wire's own summary.

← Back to the news

More stories

All the latest news