Nemotron-3-Embed-8B-BF16

Name: Nemotron-3-Embed-8B-BF16
Author: nvidia

by nvidia

8B text embedding model aimed at retrieval and RAG rather than classification

8Bparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

nvidia/Nemotron-3-Embed-8B-BF16

Feature URI

Deploy Nemotron-3-Embed-8B-BF16

Single-tenant

Mixpeek has no managed extractor for this model. On a single-tenant deployment you upload the weights and a custom plugin serves them next to the rest of your pipeline.

Read the customer-hosted guide Talk to us about single-tenant

Overview

This is a retrieval-first text embedding model. The card tags it for text-embeddings, retrieval, semantic-search and RAG, and ships it in a sentence-transformers layout with vLLM support, which tells you what it is meant to sit inside.

Size is the tradeoff to think about. At 8B parameters this is roughly two orders of magnitude larger than the MiniLM-class encoders most pipelines start with, and it will not be the thing you run over a hundred million chunks. It is the thing you run when recall on hard queries matters more than throughput, or as a reranking stage behind a cheaper first pass.

A 1B variant exists in the same family if this does not fit the budget.

Architecture

Ministral3Model, 7,952,683,008 parameters, BF16 weights. Sentence-similarity pipeline packaged for sentence-transformers and vLLM.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so Nemotron-3-Embed-8B-BF16 runs
// on your side and the output is upserted through POST
// /v1/namespaces/{namespace_id}/documents/upsert. On Enterprise the other
// path is to upload the weights instead: POST /v1/namespaces/{id}/models
// accepts the huggingface format and a custom plugin loads them.
const res = await fetch(
  "https://api.mixpeek.com/v1/namespaces/ns_your_namespace/documents/upsert",
  {
    method: "POST",
    headers: {
      Authorization: "Bearer API_KEY",
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      collection_id: "col_your_collection",
      documents: [
        {
          document_id: "asset-00412",
          // The vector name has to match a vector index on the collection.
          vectors: { "text-embedding": yourVector },
          payload: { source_key: "archive/2026/asset-00412" },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// text_extractor@v1 runs intfloat/multilingual-e5-large-instruct
// (1024-d) over a bucket, with no inference of your own.