Skip to content
#29B+Aug 8, 2026Evidence A

MiniCPM-V 4.7

ModelBest · 35B total / 3B active (MoE) + 1B edge tier

The veteran of open edge multimodal models

ModelBest’s August 2026 MiniCPM-V update, with a 35B-A3B main tier plus a 1B edge tier under Apache-2.0: documents 88, Chinese 88, ecosystem 92 and a composite of 81.8. Video 76, agents 72 and generation 70 show it remains an edge image-text assistant, not a cloud multimodal workhorse.

TextImageVideoNative

Last updated Oct 21, 2026Editorial scoring

Bottom line

For an edge Chinese image-text assistant, offline OCR or phone-gallery understanding, MiniCPM-V 4.7 is one of the most mature open options: a 1B edge tier, Apache-2.0, Chinese 88. Do not use it for cloud long video, complex agents or creative generation; its advantage is fitting, running and clean licensing.

Eight-dimension breakdown

editorial mapping · 0–100
Perception86Document88Chart QA80Video76Agent72Generation70Chinese88Ecosystem928D
  • MiniCPM-V 4.781.8
Visual perception & grounding20%
86
Document parsing & Chinese OCR18%
88
Chart & document reasoning14%
80
Video & long-horizon reasoning14%
76
GUI agents & tool use14%
72
Generation & editing6%
70
Chinese context & culture6%
88
Ecosystem, cost & availability8%
92

Weights: perception 20 / documents 18 / charts 14 / video 14 / agents 14 / generation 6 / Chinese 6 / ecosystem 8. Definitions and sources on the methodology page.

Public benchmarks

as reported, not converted
Benchmark / metricValueNote
参数规模35B-A3B / 1BMain and edge tiers
许可证Apache-2.0Commercial use
上下文128KOfficial spec

Public scores come from different versions and prompts and are not interchangeable with our composite; they are one evidence input.

Scenario mapping

five stress tests, one scale
ScenarioScoreObservation
Chinese receipt OCR88 · Stable on print
Chart data extraction80 · Fine on simple charts
30-min video QA76 · Better on short video
Mobile GUI control70 · Fine for simple tasks
Chinese cultural context88 · Good everyday context

Strengths and limitations

Strengths

  • A 1B edge tier plus a 35B-A3B main tier in one family.
  • Chinese 88 and documents 88 are enough for offline receipts and gallery understanding.
  • Ecosystem 92: model cards, quantisation scripts and edge tooling are complete.

Limitations

  • Video 76 and agents 72: long video and automation are not strengths.
  • Generation 70: not for image creation.
  • Complex charts and handwritten receipts still lose points.

Price and real cost

List price

—N/A

Open weights; edge-deployable. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥0.0 (assuming input at 25% of output price).

Use your volume

About MiniCPM-V 4.7

Is MiniCPM-V 4.7 open source?

Licence: Open weights (Apache-2.0). Weights are available, but check the exact terms and filing requirements before commercial use.

How much does MiniCPM-V 4.7 cost?

Open weights; edge-deployable. It ranks #29 with a composite of 81.8. Real bills depend on resolution, tiling, caching and volume discounts.

What is MiniCPM-V 4.7 best for?

Edge Chinese image-text assistant; Offline OCR and gallery understanding; Low-power device deployment。

Why does MiniCPM-V 4.7 score 81.8?

The composite weights eight dimensions: perception 20%, documents 18%, charts 14%, video 14%, agents 14%, generation 6%, Chinese 6%, ecosystem 8%. Its strongest dimension is Ecosystem, cost & availability; its weakest is Generation & editing.

Personal ranking · Not an official score

Every ranking, score and comparison on this page is an independent personal ranking, not an official score, and represents no vendor, institution or regulator. Scores are a subjective mapping of public evidence under our own weights — change the weights and the order changes. Disagree, or spotted a factual error? Contact the site owner: jizhi2024@foxmail.com — we check each report and log corrections in the changelog.