Pinecone Adds Gemini Embedding 2 to Managed Document Search

Pinecone announced on November 12, 2026 that its Documents API now supports Google’s Gemini Embedding 2, creating embeddings on writes and search queries through Vertex AI. The managed option removes the need for a separate Google Cloud account, API key or embedding pipeline and charges $0.20 per 1 million tokens.

The integration changes where the embedding work sits. Applications send plain text to Pinecone, which invokes the model on every write and query and retries failed calls. Embedding usage is consolidated onto the Pinecone invoice rather than billed separately by Google. Pinecone also writes each document’s text and embedding together, so semantic and full-text search use the same version of that document. Embeddings are vectors that represent meaning, allowing the same field to support similarity-based retrieval and keyword search without the application generating vectors first.

Index Configuration and Task Types

Customers select gemini-embedding-2 in a text field’s embed configuration when creating an index. Pinecone’s model documentation lists 3,072 dimensions by default, with any integer from 128 to 3,072 supported, and cosine as the sole similarity metric. The model accepts up to 8,192 tokens per input. Model, dimension and metric are fixed at index creation, making them infrastructure decisions rather than parameters teams can adjust after deployment.

Task types control how the model embeds text for different workloads. General retrieval uses retrieval_document for stored text and retrieval_query for searches by default. Classification, clustering and sentence-similarity tasks treat stored text and queries alike for text comparisons, while other query settings cover question answering, fact verification and code retrieval. A single search can override the field’s default query task.

Pinecone inserts Gemini’s required task instruction automatically, so users submit plain text instead of constructing model-specific prompts. The integrated-embedding documentation also fixes a field’s write task when the index is created. These instructions add about six or seven tokens to each value or query and count toward the model’s input limit.

Pricing, Plans and Current Limits

Charges apply to embedding tokens on both writes and queries, tying consumption to indexing and search activity. Pinecone’s Builder plan normally includes 10 million Gemini Embedding 2 tokens per month, doubled to 20 million through December 31, 2026. The model requires a Builder, Standard or Enterprise plan and is unavailable on Starter. Embedding requests can contain up to 96 inputs, while a search query is limited to 10,000 bytes.

The first release is limited to serverless Documents API indexes, excluding BYOC, pod-based and Records API indexes. Pinecone’s announcement lists console and Python SDK 10.3.0 support, with Node.js and Go SDK support coming soon. Existing vector indexes can be converted to Documents API indexes without re-embedding, but adding integrated embedding to a converted index is not yet available, so the current workflow uses new Documents API indexes. The announcement also lists multimodal support as planned for a later release.

That roadmap points beyond the text-only integration available at launch. Google’s March 10, 2026 model announcement describes Gemini Embedding 2 as natively multimodal, mapping text, images, video, audio and documents into one embedding space. Pinecone’s text service supports more than 100 languages. Processing is global, so Google can handle each request in any of its regions without guaranteeing a particular region.

For teams operating search infrastructure, the release shifts embedding from a separately managed provider integration into the database path. Its immediate fit is new text-search deployments on paid plans, with global-only processing and the converted-index limitation defining the scope of the initial service.