February Release: Performance at Scale, Predictability, and Control

The latest version of Pinecone significantly enhances the performance, predictability, and control of vector search applications, making low-latency vector search at scale more accessible for engineering teams. This update notably reduces search latency for large indexes, with benchmarks showing up to 3.4x faster search speeds for 100 million vectors. The release introduces a redesigned user console and new deployment options across GCP and AWS, offering more predictability and control while minimizing overhead. Additionally, the update improves data ingestion speed and reliability, ensuring that upserts remain fast even as the index approaches capacity. Users now have the ability to choose between different cloud providers and regions, broadening the flexibility and compliance options for Pinecone users. The improvements stem from extensive engineering efforts, including rewriting parts of the engine in Rust and optimizing I/O operations, with further enhancements anticipated. Existing indexes will automatically update by March 15, but users can request earlier updates or create new indexes with the upgraded features immediately.