The feature lets applications write raw text and submit text queries while Pinecone creates embeddings inside the service. The same text field can support semantic retrieval, sparse retrieval, and BM25 full-text search.
Three models are available at launch: llama-text-embed-v2, multilingual-e5-large, and pinecone-sparse-english-v0. The release is available on serverless indexes through Pinecone API version 2026-07, Python SDK 10.1.0, TypeScript SDK 9.1.0, and Go SDK 7.1.0 or newer. Text changes trigger re-embedding in the update; changes to other fields do not consume embedding tokens. A failed embedding causes the write not to be stored.
Pinecone says a test from a US East client at 20 queries per second recorded 26 ms median latency for the combined request using llama-text-embed-v2, versus 92 ms when a standalone embedding API call preceded search. It also says embeddings and search are billed together, but the announcement does not provide prices. The source does not establish independent latency, retrieval-quality, or cost results across workloads.