Pinecone Bring Your Own Cloud (BYOC) is generally available. Toyota Motor North America, one of the first BYOC customers, uses Pinecone to ground its AI apps in decades of manufacturing knowledge, and that knowledge never leaves Toyota's environment.
Also in this edition: on-demand indexes for BYOC, an open-source quantization benchmark, and hybrid search for Postgres apps.
Product
Pinecone BYOC Is Generally Available
With Pinecone BYOC, the data plane runs inside your own AWS, Google Cloud, or Azure account, so your vectors, documents, metadata, and request payloads never leave your boundary.
You still get the managed Pinecone experience. Pinecone handles upgrades, scaling, and maintenance without SSH, VPN, inbound network access, or a standing cross-account IAM role, and you keep using the same APIs and SDKs.
Pinecone BYOC now supports on-demand indexes alongside dedicated read nodes. Until now, every BYOC index ran on dedicated read nodes, and you chose shards and replicas for each one before creating it. On-demand indexes share one pool of nodes instead, so you can create as many as you need without sizing anything up front.
Point your app at a namespace alias and you can swap the data behind it without a redeploy: load a new namespace in the same index, check it, then repoint the alias in one atomic call. You can manage aliases in the console or through the REST API.
Go SDK v7.0.0 adds the Documents API, bringing full-text search to Go. To use it, create an index with a document schema. It's a major release with breaking changes, so check the migration guide before upgrading.
Enterprise organizations can now pull usage and cost data from the Admin API instead of going through the console. Export any period from the last 365 days in FOCUS format, the open billing format most FinOps tools already read.
Dozens of vector quantization methods have been published, each tested in its own way, so it's hard to tell which is best. We broke them into the few building blocks most of them share, released those as the open-source VQ-bench library, and benchmarked 14 quantizers side by side.
This guide walks through adding hybrid search to a Postgres app, using an online snack shop as the example. Shoppers can type a craving like "something spicy and crunchy" or an ingredient like "cheddar", and one Pinecone index handles both kinds of search through the Documents API.
Pinecone Assistant now supports Gemini 3.5 Flash for chat. With Google deprecating Gemini 2.5 Pro and OpenAI deprecating o4-mini, Assistant automatically routes those requests to Gemini 3.5 Flash and GPT-5 respectively, at the same price and with no code changes.
Restoring a backup into a different region adds a per-GB transfer fee on top of the standard restore price of $0.15 per GB. Restores within the same region are billed as before.
Pinecone Local, the in-memory Docker emulator for local testing, reached end of life on October 7. To develop and test, use a separate project in your Pinecone account, or mock the Pinecone API.
Connect with our community building with vector databases, RAG, and AI. Get help, share projects, discuss best practices, and stay updated on the latest in semantic search and retrieval-augmented generation.
