Skip to main content
Pinecone Nexus uses models at several points, and you choose them based on your deployment. You bring your own models in BYOC, or use a Pinecone-managed catalog during a Nexus trial. Everything on this page applies to both.

Model roles

  • Generation: Composes the answer to a query, using a generation model from the catalog.
  • Curation: Builds artifacts from your sources, using the Curation default and your embedding model.
  • Embedding and rerank: Power retrieval, using Pinecone-hosted models by default.
All are configurable through the model catalog, which is backed by LiteLLM and references each model by its provider/model identifier. In BYOC, you supply the keys, so see Deploy Nexus BYOC.

Default models

Each chat model call belongs to a work area, and each work area has a default model:
  • Query: Answers queries, and also runs the Optimize workflow, which tunes the manifest from query traffic.
  • Curation: Builds artifacts from your sources.
  • System: Runs the background workflows that maintain a context, such as Explore, which recommends starter manifest templates that fit your sources.
A call that doesnโ€™t name a model runs on its work areaโ€™s default, so generation runs on the Query default unless a query names a model.
In a Nexus trial, Pinecone sets the defaults from its managed catalog:
  • Query: Claude Sonnet 5.5
  • Curation: Gemini 3.8 Flash
  • System: Claude Sonnet 5.5
Rerank defaults to cohere-rerank-4-fast if your Pinecone plan includes it, and to bge-reranker-v2-m3 otherwise. Starter and Builder plans donโ€™t include cohere-rerank-4-fast (see rate limits).

Choose a model per query

A query can run on any model enabled for Query instead of the default. Name a model by its catalog ID in the queryโ€™s model field. GET /models lists the catalog IDs and the current defaults. Match the model to the question rather than the size of the context. A cheaper, faster model suits simple or high-volume queries, and the most capable model suits complex, multi-step reasoning. The query trace reports the tokens and latency behind each answer, so you can compare models on your own questions and settle on the one that fits.

Curation models

Curation runs your embedding model and the Curation default across your sources to build artifacts. Because it processes every source, curation is the most cost-sensitive step, so choose the Curation default with cost in mind. See Data flows and residency for what content each model sees.

Configure models in BYOC

In your own deployment, you configure the models yourself. You choose the default model for each work area, and the shipped catalog uses Google Gemini. You can bring any model a LiteLLM-supported provider offers, and point generation, embedding, or rerank at a different provider or an endpoint inside your own boundary, each with its own key. A chat model must support the OpenAI tools and response_format parameters, because Nexus relies on both. Align to your cloud provider where it helps:
  • Azure deployments typically use Azure OpenAI or AI Foundry models.
  • AWS deployments typically use Amazon Bedrock models.
You donโ€™t have to stay within one provider. Because the available models depend on your deployment and the keys you provide, model selection for BYOC happens with Pinecone during setup rather than from a fixed published list.

Bedrock on AWS

Bedrock models donโ€™t need a key. A Bedrock model can sign its calls with the inference proxyโ€™s own AWS identity, an IAM role attached to the proxyโ€™s Kubernetes service account, so no long-lived AWS secret sits in your model configuration. This role is attached during BYOC setup and needs bedrock:InvokeModel* for generation and embedding models and bedrock:Rerank for rerank models. Name a Bedrock rerank model by its full foundation-model ARN, for example bedrock/arn:aws:bedrock:<region>::foundation-model/cohere.rerank-v3-5:0.

Set default models

Your model catalog sets the default model for each work area in phase_defaults, and the catalog reapplies these defaults on every settings pull. To change a default, edit the catalog and redeploy. A default must be a model thatโ€™s enabled for its work area. GET /models lists the current defaults. Usage reports spend by stage, and calls on the System model appear under the system stage.