MiniCPM-V-4_5

Name: MiniCPM-V-4_5
Author: openbmb

by openbmb

Best sub-30B vision-language model with 10FPS video understanding

116Kdl/month

8Bparams

HuggingFace Run it on your own data, free

Identifiers

Model ID

openbmb/MiniCPM-V-4_5

Feature URI

mixpeek://image_extractor@v1/openbmb_minicpm_v45_v1

Overview

MiniCPM-V 4.5 is an 8B-parameter vision-language model that achieves 77.0 on OpenCompass, surpassing GPT-4o and models 10x its size. Built on Qwen3-8B with SigLIP2-400M as the vision encoder, it processes images and video with a 96x video token compression scheme that enables understanding video at 10 frames per second -- fast enough for near-real-time scene captioning.

The model excels at detailed scene description, OCR, chart understanding, and multi-image reasoning, making it a strong choice for video decomposition pipelines where each scene needs a rich caption.

Architecture

Qwen3-8B language model + SigLIP2-400M vision encoder. 96x video token compression enables 10FPS video processing. Supports multiple images and video frames in a single forward pass.

Mixpeek SDK Integration

// No extractor parameter takes a Hugging Face model id (checked against
// GET /v1/discovery/extractors, which returns 13), so MiniCPM-V-4_5 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 model produces text, so it lands in payload. Give the
          // collection a text vector index and embed that text to make it
          // searchable rather than only filterable.
          payload: { extracted_text: modelOutput, source_key: "archive/2026/asset-00412" },
          vectors: { "multimodal-embedding": embeddingOfModelOutput },
        },
      ],
    }),
  },
);

// Managed alternative, if this exact model is not the requirement:
// universal_extractor@v1 runs google/gemini-embedding-2
// (3072-d) over a bucket, with no inference of your own.