mxbai-colbert-large-v1

Name: mxbai-colbert-large-v1
Author: mixedbread-ai

by mixedbread-ai

Late-interaction ColBERT model for high-recall token-level retrieval

22Kdl/month

335Mparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

mixedbread-ai/mxbai-colbert-large-v1

Feature URI

mixpeek://text_extractor@v1/mxbai_colbert_large_v1

Overview

mxbai-colbert-large-v1 from mixedbread.ai is a ColBERT-style late-interaction retrieval model that produces per-token embeddings for both queries and documents. Instead of compressing an entire passage into a single vector, it retains token-level representations and computes relevance via MaxSim: the maximum similarity between each query token and all document tokens. This architecture captures fine-grained lexical and semantic matches that single-vector models miss.

Architecture

Late-interaction transformer based on the ColBERT architecture. Encodes queries and documents independently into per-token embeddings, then scores relevance using MaxSim: for each query token, find the maximum cosine similarity with any document token, then sum across query tokens. This allows pre-computation of document embeddings while retaining token-level matching at query time.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so mxbai-colbert-large-v1 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.