We're rolling out new features every day during AWS re:Invent.
Be the first to try them - watch this space or subscribe for updates.
Seamlessly combine dense retrieval, sparse retrieval, and reranking into a unified search pipeline, delivering up to 48% better performance — and 24% better, on average — than either sparse or dense alone.
Our sparse embedding model is designed for high-precision and efficient retrieval — outperforming BM25 by up to 44%. Fully integrated into Pinecone’s infrastructure, it streamlines the development of performant search applications with state-of-the-art sparse retrieval capabilities.
DocumentationThis new index type enables direct indexing and retrieval of sparse vectors, supporting traditional methods like BM25 and advanced learned sparse models like pinecone-sparse-english-v0. By separating dense and sparse workflows, you gain greater control over retrieval strategies, optimizing results for specific use cases.
Request Early AccessWe introduced embedding and reranking earlier this year. Today, we’re announcing the GA of those capabilities and bringing inference closer to the database with a native integration and fully managed support for new models developed by Pinecone and Cohere so you can now embed, rerank, and query your data with a single API.
Streamline AI development with fast, easy access to leading embedding and reranking models via new endpoints that integrate embedding and reranking.
DocumentationWhether you're looking to improve internal search capabilities or strengthen your RAG pipelines, our new reranker — pinecone-rerank-v0 — is built to meet the demands of modern, large-scale applications while delivering top-tier performance.
Our research shows pinecone-rerank-v0 improves search accuracy by up to 60% and on average 9% over industry-leading models on the BEIR benchmark. Try it out today via the Inference API!
DocumentationEasily select and use cohere-rerank-v3.5 directly from the Pinecone API to enhance the relevance of your search results. Rerank 3.5 excels at understanding complex business information across languages making it optimal for global organizations in sectors like finance, healthcare, the public sector, and more.
DocumentationWe've expanded our security suite with support for CMEK and RBAC with API key roles. Both features are now available in public preview. We've also announced audit logs in early access, and the GA of Private Endpoints for AWS PrivateLink.
CMEK allows you to encrypt your data using keys that you manage in your cloud provider's key management system (KMS). Pinecone currently supports CMEK using AWS KMS.
DocumentationAPI key roles create a more comprehensive access control system that helps you mitigate security risks, streamline operations, and manage resources more efficiently. This new functionality includes six roles with varying permissions across control and data planes.
DocumentationAdd an additional layer of security to your serverless index with Private Endpoints for AWS PrivateLink, which is now generally available (GA).
DocumentationAudit logs provide a detailed record of user and API actions that occur within the Pinecone platform.
DocumentationYou can now prototype, test, and develop AI applications faster with Pinecone Local, a self-hosted, in-memory emulator of the vector database. It is now available in public preview for all users.
Pinecone Local is available via Docker images: The pinecone-local image provides the full vector database emulator, which enables you to add/delete indexes using our API to build out your environment and run your full suite of tests. The pinecone-index image lets you spin up a single local Pinecone index without starting the full emulator.
Documentation