Perplexity has introduced Photon, an in-house retrieval and ranking engine built in Rust, replacing the company’s earlier fork of an open-source engine. The company said Photon now handles all production traffic and powers the new "fast search" mode in its search API. Reported latency figures for a single call were 160 ms at p50 and 230 ms at p95, while p99 latency for retrieval and ranking dropped from about 800 ms to about 65 ms. Perplexity also said Photon uses roughly 20% fewer service machines than its older content nodes. The engine is available as a hosted API through POST /search with search_type set to "fast," priced at $1 per 1,000 requests, though the engine itself is not open-source. Alongside the launch, Perplexity released a "fast search" mode tuned for agent workflows. In 3,554 tasks across six benchmarks, that mode scored 64.3% with an estimated combined model and search cost of $59.73, compared with a default preset cost of $187.60, making it about 68% cheaper, according to MarkTechPost.
Perplexity has released Photon, an in-house retrieval and ranking engine written in Rust, as it moves away from its earlier fork of an open-source engine.
The company said Photon now serves all production traffic and supports the new "fast search" mode in its search API. Perplexity reported single-call latency of 160 ms at p50 and 230 ms at p95.
Latency and infrastructure figures
According to Perplexity, Photon reduced p99 latency for retrieval and ranking from about 800 ms to about 65 ms. The company also said the engine uses roughly 20% fewer service machines than its previous content nodes.
API access and pricing
Photon is offered as a hosted API. Users can access it by setting search_type to "fast" in POST /search. Pricing is listed at $1 per 1,000 requests, while the engine itself is not open-source.
Fast search benchmark results
Perplexity also introduced a "fast search" mode designed for agent workflows alongside Photon. In 3,554 tasks across six benchmarks, the mode posted a score of 64.3% with an estimated combined model and search cost of $59.73. The default preset cost was listed at $187.60, making the new mode about 68% cheaper.
The report was cited by MarkTechPost.
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
