Tools & ProjectsSeptember 30, 2026
17
š±
Perplexity's Rust Engine Crushes Latency
Perplexity introduces Photon, a Rust-based retrieval engine that cuts p99 latency from 800 ms to 65 ms.
#Perplexity#Rust#Search#Infrastructure#Latency
š„ What happened
Perplexity launched Photon, an in-house Rust retrieval and ranking engine that replaces its old open-source fork. It now serves all production traffic, cutting p99 latency from roughly 800 ms to about 65 ms.
š” Why it matters
That's a 12x latency drop on 20% fewer serving machines, storing 2.5x more data per document. The new Fast Search mode costs $1 per 1,000 requests and cuts estimated agent task cost by about 68% at comparable quality.
ā” Our take
If you're still maintaining a forked search engine, Perplexity just proved the fork was more expensive than building from scratch. Time to reconsider your stack.
The title, summary and analysis of this item were produced automatically by an AI system and have not been editorially reviewed. They may contain errors, bias or omissions ā when in doubt, read the linked original source.
