Perplexity has deployed Photon, an in-house retrieval and ranking engine written in Rust, across its production search infrastructure. The system replaces a previously forked open-source engine and now handles retrieval and ranking for all production traffic, reducing p99 latency from 800 milliseconds to 65 milliseconds according to the company. Photon powers a new Fast Search mode in Perplexity’s product (MarkTechPost).
Liquid AI released d1, a decision model designed for structured choice problems rather than text generation. The model accepts context and a set of typed questions, then returns calibrated probabilities across a fixed set of outcomes in a single inference call with zero generated tokens, targeting workflows where many teams currently use text-generation models (MarkTechPost). Separately, NVIDIA researchers introduced Physis-Lang, a framework that uses learned language representations to improve physics accuracy in video world models, lifting Cosmos 3 past Veo 3.1 on physics benchmarks (MarkTechPost).